Method and system for combinatorial optimization

Abstract

The present invention relates to a method and a system for combinatorial optimization. More particularly, the present disclosure relates to a method and a system for combinatorial optimization, in which parameter optimization is performed through a combination of a hybrid evolutionary algorithm (EA) and Bayesian optimization (BO).

Description

Combination Optimization Method and System

[0001] The present invention relates to a combinatorial optimization method and system. More specifically, the present invention relates to a combinatorial optimization method and system that performs parameter optimization through a combination of a hybrid evolutionary algorithm (EA) and Bayesian optimization (BO).

[0002] Optimization problems aim to find the best solution under given conditions. Such optimization problems play a crucial role in various industries and research fields, including manufacturing, telecommunications, finance, and artificial intelligence (AI). In particular, efficiently exploring complex and multidimensional parameter spaces is essential in diverse areas, such as manufacturing processes, parameter tuning for machine learning models, and performance enhancement for image processing systems.

[0003] In this regard, various industrial sectors frequently face the challenge of optimally adjusting numerous setting variables (or parameters) under complex conditions. For example, in visual inspection systems used in manufacturing processes, multiple variables such as lighting intensity, camera position, resolution, and image processing speed interact with each other, and inspection accuracy and processing speed vary significantly depending on their combination.

[0004] In such an environment, testing every possible combination of variables is inefficient in terms of time and cost, and the resources required to find the optimal combination increase, especially as the number of variables grows. Furthermore, since even minute changes to some variables can significantly impact overall performance, techniques are needed to efficiently explore and precisely tune from the initial setup phase.

[0005] In this regard, various algorithms to solve optimization problems are being actively researched recently. However, conventional algorithms may have excellent global search capabilities but are vulnerable to local optimization, or they may experience performance degradation when initial data is insufficient, even if precise local search is possible.

[0006] As discussed above, in fields requiring the consideration of interactions among multiple parameters (e.g., visual inspection systems), it is often difficult to quickly derive the optimal solution using only a single algorithm. Therefore, there is a need for an optimization method that overcomes the limitations of conventional algorithms and ensures both initial search speed and the accuracy of the final solution.

[0007] The present invention is intended to provide a combination optimization method and system capable of solving optimization problems and improving optimization performance.

[0008] More specifically, the present invention is intended to provide a combinatorial optimization method and system capable of extensively exploring the initial solution space and performing efficient and accurate parameter optimization.

[0009] Furthermore, the present invention aims to provide a combinatorial optimization method and system capable of ensuring both search speed and the accuracy of the final solution, even in complex environments where multiple parameters interact.

[0010] Furthermore, the present invention aims to provide a combinatorial optimization method and system capable of minimizing resource consumption during the optimization process and calculating an optimal solution that considers the interaction of multiple parameters.

[0011] To solve the problem described above, a computerized combinatorial optimization method according to the present invention, comprising the following steps, may include: specifying an input value related to a parameter to be optimized; performing a search for a plurality of candidate solutions generated according to the input value using a pre-set first optimization algorithm; generating result data based on the results of the search for the plurality of candidate solutions; training an artificial intelligence model using the generated result data; and performing optimization for the parameter to be optimized based on the trained artificial intelligence model and a pre-set second optimization algorithm.

[0012] In an embodiment, the input value includes at least one of the number of candidate solutions including the optimization target parameter, the number of optimization target parameters included in the candidate solutions, and the search range of the optimization target parameter, and the step of performing the search is a step of repeatedly performing the search using the first optimization algorithm based on a preset number of iterations for the first optimization algorithm, and in the step of generating the result data, the result data can be generated based on the search results repeatedly performed according to the preset number of iterations.

[0013] In an embodiment, in the step of performing the search, the search using the first optimization algorithm is performed N times, corresponding to the number of repetitions N set, and in the step of generating the result data, the result data can be generated based on the search results performed N times.

[0014] In an embodiment, in the step of performing the search, the plurality of candidate solutions are generated according to the input value using the first optimization algorithm, and a search is performed for each of the plurality of candidate solutions to select an optimal candidate solution that satisfies a preset criterion from the plurality of candidate solutions, and in the step of generating the result data, the result data can be generated based on the results of the search performed for each of the plurality of candidate solutions.

[0015] In an embodiment, the step of generating the plurality of candidate solutions may generate the plurality of candidate solutions based on the search range of the optimization target parameter included in the input value.

[0016] In an embodiment, the step of performing the search involves performing an evaluation for each of the plurality of candidate solutions to produce an evaluation result for each of the plurality of candidate solutions, and based on the evaluation result for each of the plurality of candidate solutions, selecting the optimal candidate solution among the plurality of candidate solutions that satisfies the preset criteria, and the result data may include at least one of the plurality of candidate solutions and the evaluation result for each of the plurality of candidate solutions.

[0017] In an embodiment, at least some of the plurality of candidate solutions can be updated using a preset operation method, and a search for the updated candidate solutions can be performed to select an optimal candidate solution that satisfies the preset criteria from the updated candidate solutions.

[0018] In an embodiment, an evaluation of the updated candidate solution is performed to produce an evaluation result for the updated candidate solution, and based on the evaluation result for the updated candidate solution, the optimal candidate solution among the updated candidate solutions that satisfies the preset criteria is selected, and the result data may further include at least one of the updated candidate solution and the evaluation result for the updated candidate solution.

[0019] In an embodiment, the step of performing the search is to search for at least one optimal candidate solution for each iteration based on a preset number of iterations for the first optimization algorithm, and the optimal candidate solution can be included in the result data and used for training the artificial intelligence model.

[0020] In an embodiment, the result data may include at least one of the optimal candidate solution, the evaluation result for the optimal candidate solution, and the search range of the optimal candidate solution.

[0021] In an embodiment, to train the artificial intelligence model, the step of storing the result data in a specified storage is further included, and in the step of performing the optimization, optimization can be performed on a search space defined based on the optimization target parameter.

[0022] In an embodiment, in the step of training the artificial intelligence model, the artificial intelligence model is trained using the result data stored in the specified repository to perform optimization on the optimization target parameter, and in the step of performing the optimization, optimization on the search space can be performed based on the trained artificial intelligence model.

[0023] In the embodiment, in the step of performing the optimization, optimization of the target parameter can be performed through at least one parameter combination specified from the search space using the learned artificial intelligence model and a specific function.

[0024] In an embodiment, the learned artificial intelligence model may be configured to predict at least one of a predicted value and a predicted variance for at least one combination of parameters.

[0025] In an embodiment, in the step of performing the optimization, the learned artificial intelligence model may be updated based on the parameter combination specified from the search space, and optimization of the target parameter may be performed based on the update to the learned artificial intelligence model.

[0026] In the embodiment, in the step of performing the optimization, optimization of the target parameter for optimization is performed based on the updated artificial intelligence model and the second optimization algorithm to produce an optimal solution for the target parameter for optimization.

[0027] In an embodiment, the method further includes a step of estimating a prior distribution for the optimization target parameter using a pre-set probability distribution estimation technique, and in the step of training the artificial intelligence model, the artificial intelligence model can be trained using at least one of the prior distribution for the optimization target parameter and the result data.

[0028] In an embodiment, the optimization target parameter includes a plurality of parameters having different characteristics, and in the step of estimating the prior distribution, the prior distribution for each of the plurality of parameters is estimated using the pre-set probability distribution estimation technique, and in the step of training the artificial intelligence model, the artificial intelligence model can be trained using at least one of the prior distribution for each of the plurality of parameters and the result data.

[0029] A combinatorial optimization system according to the present invention, comprising a memory configured to store executable instructions and one or more processors configured to perform operations by executing one or more instructions, can specify input values ​​related to parameters to be optimized, perform a search for a plurality of candidate solutions generated according to the input values ​​using a pre-configured first optimization algorithm, generate result data based on the results of the search for the plurality of candidate solutions, train an artificial intelligence model using the generated result data, and perform optimization for the parameters to be optimized based on the trained artificial intelligence model and a pre-configured second optimization algorithm.

[0030] A program according to the present invention is a program that is executed by one or more processes in an electronic device and can be stored on a computer-readable recording medium, and may include instructions for performing steps of: specifying an input value related to a parameter to be optimized; performing a search for a plurality of candidate solutions generated according to the input value using a first optimization algorithm set in advance; generating result data based on the results of the search for the plurality of candidate solutions; training an artificial intelligence model using the generated result data; and performing optimization for the parameter to be optimized based on the trained artificial intelligence model and a second optimization algorithm set in advance.

[0031] As described above, according to the combinatorial optimization method and system of the present invention, by combining the global search capability of an evolutionary algorithm (EA) and the precise local search capability of Bayesian optimization (BO), it is possible to extensively explore the initial solution space and perform fine parameter adjustments. Accordingly, the present invention effectively prevents the cold start problem, in which the prediction or search performance of a model (or algorithm) deteriorates due to a lack of initial data, and can derive the optimal solution to the problem quickly and accurately. In other words, the present invention can achieve stable and high-performance optimization even in situations where initial data is insufficient.

[0032] Furthermore, according to the combinatorial optimization method and system of the present invention, various initial solutions are efficiently generated through the search of an evolutionary algorithm and utilized in Bayesian optimization, thereby enabling efficient and accurate parameter optimization even with limited data. Through this, the present invention can maintain high optimization performance while minimizing the consumption of computational (or operation) resources, and can effectively search a complex multidimensional parameter space.

[0033] Furthermore, according to the combinatorial optimization method and system of the present invention, for a plurality of interacting parameters (or variables), optimization reflecting the interaction effects between multiple parameters is possible by learning and adjusting candidate combinations generated using an evolutionary algorithm through Bayesian optimization. Through this, the present invention can simultaneously improve the speed and accuracy of optimization even in environments requiring real-time data, such as parameter tuning of machine learning models, industrial process optimization, and visual inspection machines. Accordingly, the present invention can provide efficient and accurate optimization results in various fields such as parameter tuning of machine learning models, industrial process optimization, and visual inspection machines. Consequently, the present invention converges to an optimal solution faster and more stably than when using a single algorithm alone, thereby providing excellent performance and resource efficiency across complex combinatorial optimization problems and enabling the advancement of parameter combination optimization.

[0034] Meanwhile, according to the combinatorial optimization method and system of the present invention, optimization performance can be improved by estimating the individual distribution for each of the multiple parameters in an environment where multiple parameters having different characteristics are mixed. That is, by processing each parameter individually, the present invention can prevent performance degradation in high-dimensional problems and mitigate the curse of dimensionality.

[0035] Furthermore, according to the combinatorial optimization method and system of the present invention, an artificial intelligence model can be trained based on at least one of the result data generated using a prior distribution and an evolution algorithm for each of a plurality of parameters, and parameter optimization can be performed based on the trained artificial intelligence model. Through this, the present invention can minimize the total search space and concentrate search resources on a valid and high-performance space, thereby reducing unnecessary evaluations and improving the optimization convergence speed.

[0036] Furthermore, according to the combination optimization method and system of the present invention, computational costs are reduced by estimating the prior distribution for each of a plurality of parameters having different characteristics, and individual learning rates or bandwidth adjustments are possible to suit the characteristics of each parameter.

[0037] Furthermore, according to the combination optimization method and system of the present invention, each parameter can be processed simultaneously according to the characteristics of each of the multiple parameters, and accurate estimation is possible even within a wide optimization range of each parameter.

[0038] Furthermore, according to the combinatorial optimization method and system of the present invention, when only some parameters among a plurality of parameters having different characteristics are changed, only the prior distribution of the changed parameters can be selectively updated. Through this, the present invention can improve the efficiency of incremental learning and can apply different bandwidths to each parameter, thereby increasing the accuracy of distribution estimation according to the characteristics of each variable. In addition, distribution information of some parameters can be reused from similar optimization problems to be quickly applied to new problems, and the distribution of related parameters can be immediately updated whenever new data is received, thereby enabling real-time response to environmental changes. Accordingly, since the present invention only requires storing the distribution information of each parameter, memory can be used efficiently to save storage space.

[0039] As such, the present invention efficiently processes complex mixed variables and enables accurate estimation even within a wide optimization range for each variable. This method can provide improved performance in high-dimensional problems, effective processing of mixed variables, and increased computational efficiency. In particular, the present invention can be usefully applied to real industrial environments where interactions between variables are complex and the range of influence of each variable is wide, such as factory automation or complex system optimization.

[0040] FIG. 1 is a conceptual diagram illustrating a combination optimization system according to the present invention.

[0041] FIGS. 2, FIGS. 3 and FIGS. 4 are flowcharts illustrating a combination optimization method according to the present invention.

[0042] FIGS. 5 and FIGS. 6 are formulas related to the combination optimization method according to the present invention.

[0043] FIG. 7 shows an example of an algorithm of a combination optimization method according to the present invention.

[0044] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components are assigned the same reference number regardless of the drawing symbols, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not have distinct meanings or roles in themselves. Furthermore, in describing the embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.

[0045] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0046] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0047] A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0048] In this application, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0049] The present invention relates to a combinatorial optimization method and system capable of solving optimization problems and improving optimization performance. More specifically, the present invention relates to a combinatorial optimization method and system capable of extensively exploring an initial solution space and performing efficient and accurate parameter optimization. The combinatorial optimization method and system according to the present invention may also be referred to as a “parameter optimization method and system.”

[0050] Furthermore, the present invention relates to a combinatorial optimization method and system capable of performing Bayesian optimization in an environment where continuous and discrete variables (or parameters) are mixed. More specifically, the present invention relates to a combinatorial optimization method and system that operates effectively even in a high-dimensional environment (or mixed-variable environment) where continuous and discrete variables are mixed.

[0051] A mixed-variable environment refers to a variable space where continuous and discrete variables coexist and interact within the same optimization problem. For example, in a mixed-variable environment, the data types, ranges, and resolutions of each variable differ, making it complex or inefficient to process all variables using a single joint probability model (or module). This tendency can be exacerbated, particularly when the number of variables is large or the optimization range of each variable is wide. Such mixed-variable environments can appear in various industrial fields, including factory automation, complex system parameter tuning, and production process optimization.

[0052] Here, a continuous variable can mean a variable that can take any real value (or infinitely many values) within a given specific interval. A continuous variable has a set of values ​​that belongs to a continuous interval or a subset of the entire set of real numbers, and between two adjacent values, there can be infinitely many different values ​​(or always different values). For example, a continuous variable can have a continuity of values ​​in which there are infinitely many intermediate values ​​between two values ​​(e.g., between 1.2 and 1.3, there exist 1.25, 1.1251, 1.2511, etc.). Examples of such continuous variables may include variables that numerically express physical / quantitative characteristics, such as length (e.g., 1.2 m, 1.25 m, 1.257 m, etc.), temperature (e.g., 25.3℃, 25.31℃, 25.314℃, etc.), pressure (e.g., 1.01 MPa, 1.011 MPa, etc.), time (e.g., 1.5 sec, 1.53 sec, 1.534 sec, etc.), lighting intensity, voltage / current (e.g., 3.3V, 3.31V, 3.314V, etc.), speed (e.g., 50.2 km / h, 50.23 km / h, etc.). In this specification, the term continuous variable may be used interchangeably with “first parameter,” “continuous parameter,” “first variable,” “continuous variable,” or “continuous parameter.”

[0053] Additionally, a discrete variable may refer to a variable capable of taking only a finite or countable infinite number of distinct values. Discrete variables are expressed solely by distinct individual values ​​rather than a continuous range of real numbers, and there may be no other value (or continuous value) between the two values. Such discrete variables may include at least one of integer values ​​(e.g., production process step number, equipment mode setting, etc.), nominal values ​​(e.g., color, category, etc.), ordinal values ​​(e.g., grade, level, etc.), or boolean values ​​(e.g., ON / OFF, etc.). For example, examples of discrete variables may include process steps (e.g., {Step 1: 0.4, Step 2: 0.35, Step 3: 0.25}), algorithm types (e.g., {A: 0.5, B: 0.3, C: 0.2}), number of product defects (e.g., 0, 1, 2, ...), machine operation modes (e.g., {Standby, Running, Stopped}), quality grades (e.g., High, Medium, Low), color classifications (e.g., {Red, Blue, Green}), production line numbers (e.g., No. 1, No. 2, No. 3...), part specification codes (e.g., {A1, A2, B1, B2}), etc. In this specification, the term discrete variable may be used interchangeably with “second parameter,” “discrete parameter,” “second variable,” “discrete variable,” “discrete parameter,” etc.

[0054] However, the examples included in continuous and discrete variables are not necessarily limited to those mentioned above, and it is obvious that various other examples may be included in addition to those mentioned.

[0055] Furthermore, Bayesian optimization is a stochastic optimization technique for efficiently searching for the optimal value of an objective function that is costly or computationally complex. Bayesian optimization is a technique (or algorithm) that efficiently searches a given solution space using a surrogate model and / or a probabilistic model (e.g., a Gaussian Process (GP) model). Bayesian optimization can induce rapid convergence by proceeding with the search based on probabilistic predictions from given data. For example, instead of directly iteratively evaluating the objective function, Bayesian optimization can estimate the distribution of the objective function using a surrogate model and / or a probabilistic model, and select the next search point through an acquisition function. Such Bayesian optimization may be one algorithm included in the optimization algorithm, and in the present invention, it may also be referred to as a “pre-configured second optimization algorithm.”

[0056] Meanwhile, conventional combinatorial optimization algorithms suffered from the problem that search efficiency deteriorated sharply and computational resource consumption increased as the search space became more dimensional or the interdependencies between parameters became more complex. In particular, existing search techniques had limitations in that they failed to reach a global optimum and converged to a local optimum due to restricted initial solution spaces or biased search paths. Furthermore, in environments where multiple optimization parameters interacted, it was difficult to achieve overall performance improvement of the optimization target solely through iterative searches at the individual parameter level. In addition, existing methods based on static search strategies faced the problem that it was difficult to simultaneously ensure computational efficiency and result accuracy because adaptive updates based on intermediate search results were impossible.

[0057] To solve the above problem, the present invention provides a combinatorial optimization method and system capable of simultaneously improving search efficiency and convergence stability by generating result data based on initial search results using a first optimization algorithm and utilizing this for the training of an artificial intelligence model. Furthermore, by reflecting the prediction results of the trained artificial intelligence model in the search process of a second optimization algorithm, it enables the calculation of a more precise optimal solution while preventing the waste of search resources. In addition, the present invention minimizes the usage of computational resources and maintains the stability of the search process by dynamically adjusting the range of the search space or the distribution of candidate solutions. Accordingly, the present invention solves the technical problem of enabling efficient and accurate optimization while reflecting complex parameter interactions.

[0058] According to the present invention, by converting the search results generated by a first optimization algorithm into training data, learning the interaction between search patterns and parameters through an artificial intelligence model, and reflecting the predicted values ​​and predicted variances of the learned model into a second optimization algorithm, a technical effect can be obtained in which search efficiency, convergence stability, and optimal solution quality are simultaneously improved.

[0059] In particular, the present invention provides a dynamic feedback loop in which performance is cumulatively improved as the search strategy is repeated by forming a structure that cycles the entire search process into “search-learn-research.” Through this, the search time is shortened compared to existing static search techniques, and higher solution quality can be achieved within the same computational resources.

[0060] In addition, computational efficiency is improved by reducing the evaluation of unnecessary candidate solutions during the search process and efficiently managing resource usage.

[0061] Furthermore, since the artificial intelligence model directly contributes to the control of the search space and parameter adjustment, the application of artificial intelligence as a technical means beyond simple data analysis is realized. Accordingly, the present invention enables stable and accurate optimization even in high-dimensional search spaces and provides the technical effect of improving the reliability of convergence for the global optimal solution.

[0062] Combinatorial optimization as described in the present invention may refer to a mathematical or computational procedure for solving the problem of optimizing (maximizing or minimizing) a given objective function among a combination of multiple discrete variables or parameters. That is, it includes a process of searching for the optimal solution among all possible combinations of candidate solutions that satisfies constraints and yields the highest value of the objective function.

[0063] Combinatorial optimization problems generally have very large search spaces, and because exhaustive search is impractical due to their characteristics, approximate solutions are sought using heuristic or metaheuristic approaches such as genetic algorithms, simulated annealing, particle swarm optimization, and Bayesian optimization. In particular, when multiple parameters are interdependent or the objective function has a non-linear or multimodal structure, combinatorial optimization problems are difficult to solve using traditional mathematical optimization methods, and high-dimensional search techniques that simultaneously consider search efficiency and convergence stability are required. The present invention relates to a technology that enables the efficient solution of such combinatorial optimization problems through an artificial intelligence-based search and learning structure.

[0064] Meanwhile, the present invention can be usefully utilized in various situations. More specifically, the combination optimization method and system according to the present invention can be usefully utilized by applying them to various industries and services. For example, as shown in FIG. 1, the present invention can be usefully utilized by applying it to a vision inspection system that determines (or predicts) whether a product to be manufactured (20, or an image of the product to be manufactured (20)) is abnormal (e.g., normal (21) or defective (22)) when the product to be manufactured (20, or an image of the product to be manufactured (20)) is received (or input). However, the fields to which the present invention is applied are not necessarily limited to the fields mentioned above, and can be usefully applied to actual industrial environments where the interaction between variables is complex and the range of influence of each variable is wide (e.g., factory automation or complex system optimization).

[0065] Hereinafter, the present invention will be examined in more detail with reference to the attached drawings. FIG. 1 is a conceptual diagram illustrating a combination optimization system according to the present invention. FIGS. 2, 3, and 4 are flowcharts illustrating a combination optimization method according to the present invention. FIGS. 5 and 6 are formulas related to a combination optimization method according to the present invention. Furthermore, FIG. 7 shows an embodiment of an algorithm of a combination optimization method according to the present invention.

[0066] Meanwhile, as illustrated in FIG. 1, the combination optimization system (100) according to the present invention may include at least one of an input unit (110), an output unit (120), a communication unit (130), a storage unit (140), a probability distribution estimation unit (150), a data processing unit (160), an artificial intelligence model (170), and a control unit (180).

[0067] A combinatorial optimization system (100) according to the present invention may include at least one processor and at least one memory containing computer program code. In this case, the memory may function as the storage unit (140) above. In the present invention, the memory and the program code may cooperate with the processor to perform a series of processes described below. Although not illustrated, the combinatorial optimization system (100) according to the present invention may include one or more processors, and such processors may include one or more general-purpose processors and / or one or more special-purpose processors (e.g., digital signal processor, tensor processing unit (TPU), graphics processing unit (GPU), neural network processing unit (NPU), application integrated circuit, application semiconductor (ASIC), field programmable gate array (FPGA), quantum processing unit (or quantum processor, QPU), etc.). One or more processors may be configured to execute instructions, computer-readable instructions, and / or other instructions described herein that are stored (or included) in the storage unit (140). The combination optimization method and system according to the present invention can perform the data processing described below by means of a memory and at least one processor cooperating. The processor can perform a series of operations and data processing using data and information stored in the memory. At this time, the memory may be a component of the storage unit (140).

[0068] In addition, the combinatorial optimization system (100) according to the present invention can perform data processing and computation processes using quantum gates, quantum entanglement, and quantum superposition states, taking into consideration implementation in a quantum computer environment. For example, the present invention can perform parallel computations based on qubits, and such quantum computations can operate complementarily with existing classical computers.

[0069] Such quantum computers may include parallel computation using qubits and high-speed data processing devices utilizing quantum entanglement, and hardware-based computational optimization using FPGAs and ASICs is possible. In addition, quantum computers may utilize quantum processors capable of qubit-based parallel computation, and data processing efficiency can be improved through a hybrid structure with existing classical computers.

[0070] Meanwhile, the input unit (110) can be configured in various ways as a means of data input. For example, the input unit (110) can be configured to receive user input. The input unit (110) can be configured to receive user input from a user terminal (10). Here, “receiving input” may mean receiving an input signal (or selection signal) corresponding to the user’s input based on input made by the user through the input unit configuration provided in the user terminal (10).

[0071] Here, the user terminal (10) may include at least one of a mobile phone, a smartphone, a notebook computer, a laptop computer, a slate PC, a tablet PC, an ultrabook, a desktop computer, a digital broadcasting terminal, a PDA (personal digital assistants), a PMP (portable multimedia player), a navigation device, and a wearable device (e.g., a smartwatch, a smart glass, a head-mounted display).

[0072] In addition, the input unit (110) in the present invention does not necessarily mean a hardware means, but can be understood as a channel for receiving input from a user.

[0073] The input unit (110) may also be referred to as a user interface module. The input unit (110) may include a touch screen, a computer mouse, a keyboard, a keypad, a touchpad, a trackball, a joystick, a voice recognition module, or other similar devices. However, the present invention does not limit the type of input unit (110).

[0074] Here, user input may include documents, text, images (or videos), voice, etc. In this case, the combination optimization system (100) may further include a module that converts voice into text.

[0075] Next, the output unit (120) can output information through an output unit configuration (e.g., a display unit, a touch screen, a speaker, etc.) provided in a user terminal (10) linked to the combination optimization system (100) according to the present invention. For example, the output unit (120) can output a page (or service page) linked to the combination optimization system (100) according to the present invention to the display unit of the user terminal. In addition, the output unit (120) does not necessarily mean a hardware means, but can be understood as a channel for outputting results to a user.

[0076] Next, the communication unit (130) may be connected via a wireless or wired network to a user terminal, a server (e.g., a central server, an external server, etc.), a device, and at least one network, etc., to receive or transmit overall data and information necessary for the operation of the combination optimization system (100) according to the present invention.

[0077] The communication unit (130) can support various communication methods depending on the communication standard of the communicating device.

[0078] For example, the communication unit (130) may be configured to communicate with a communication target using at least one of the following technologies: WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband), WiMAX (World Interoperability for Microwave Access), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), 5G (5th Generation Mobile Telecommunication), Bluetooth (Bluetooth™ Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra-Wideband), ZigBee, NFC (Near Field Communication), Wi-Fi Direct, and Wireless USB (Wireless Universal Serial Bus).

[0079] Next, the storage unit (140, or memory) serves to store various data related to the present invention and may include one or more non-transient computer-readable storage media that can be read and / or accessed by at least one of one or more processors.

[0080] One or more computer-readable storage media may include volatile and / or non-volatile storage components, such as optical, magnetic, organic, or other memory or disk storage devices. In some examples, the storage unit (140) may be implemented using a single physical device (e.g., one optical, magnetic, organic, or other memory or disk storage device), whereas in other examples, the storage unit (140) may be implemented using two or more physical devices.

[0081] The storage unit (140) may include computer-readable instructions and additional data. The storage unit (140) may include a storage necessary to perform at least some of the methods, scenarios, and techniques described herein and / or at least some of the functions of the device and network.

[0082] Furthermore, at least a portion of the storage unit (140) may be a cloud storage or a cloud server. The storage unit (140) may store at least some of the data corresponding to user input received from the input unit (110) (e.g., an image (20) of a product to be manufactured, discrete variables and continuous variables, etc.) and training data.

[0083] That is, the storage unit (140) is sufficient as a space where information necessary for the operation of the combination optimization system (100) according to the present invention is stored, and it can be understood that there are no restrictions on the physical space.

[0084] Furthermore, the storage unit (140) may store a computer program including computer program instructions. Furthermore, the storage unit (140) may store a computer program including computer program instructions that control the operation of the system (100) or control the operation of the control unit (180) when loaded into the processor of the system (100).

[0085] Next, the probability distribution estimation unit (150) may be configured to perform the role of estimating a prior distribution for each of a plurality of parameters having different characteristics (or data characteristics) using a pre-set probability distribution estimation technique. Here, the pre-set probability distribution estimation technique (or probability density function estimation technique) may include a Kernel Density Estimation (KDE) technique.

[0086] Kernel density estimation is a non-parametric method for estimating the probability density function (PDF) of data from given data samples. It approximates the distribution of data within a specific interval as a smooth curve. For example, kernel density estimation can be performed by placing kernel functions around observed samples and summing the total density functions, without assuming the shape of the data distribution. This method smoothly represents the structure of the data and allows for the application of various kernels tailored to the characteristics of the variables (e.g., continuous or discrete variables).

[0087] The probability distribution estimation unit (150) can estimate the individual distribution (or prior distribution, probability distribution, etc.) of each parameter by performing kernel density estimation for each of a plurality of parameters having different ranges and characteristics. In this case, the probability distribution estimation unit (150) can estimate the prior distribution by applying different probability distribution estimation techniques to each of the plurality of parameters. For example, the probability distribution estimation unit (150) can model the prior distribution by applying a Gaussian kernel density estimation (Gaussian KDE) technique to the first parameters (continuous variables) and a categorical probability distribution estimation (KDE for discrete variables) technique to the second parameters (discrete variables). That is, the pre-set probability distribution estimation technique may include at least one of the Gaussian kernel density estimation technique and the categorical probability distribution estimation technique.

[0088] The Gaussian probability distribution estimation technique is a nonparametric method for estimating the probability density function of a first parameter. It involves transforming each data sample (or data point) into a normal distribution (or normal kernel) where the mean is the corresponding sample value and the variance (or standard deviation) is the bandwidth (h), and then summing all of these to estimate the overall probability density function (or approximate the overall distribution). In this specification, the Gaussian probability distribution estimation technique may also be referred to as the “pre-established first probability distribution estimation technique.”

[0089] In addition, the categorical probability distribution estimation technique is a KDE method that estimates a probability mass function based on the frequency of occurrence (or frequency of occurrence) for each category (or category) to estimate the distribution of categorical data (or second parameter). Unlike the Gaussian probability distribution estimation technique, which uses kernels in a continuous space, the categorical probability distribution estimation technique can estimate probabilities by applying a categorical kernel (e.g., Dirichlet kernel, based on a delta function) to each category. That is, unlike continuous KDEs, the kernel function of the categorical probability distribution estimation technique can be defined based on whether the categorical data is identical. For example, an index function that assigns 1 if the categories are the same and 0 if they are different, or other kernels that reflect similarity, can be used. In this specification, the categorical probability distribution estimation technique may also be referred to as a “pre-configured second probability distribution estimation technique.”

[0090] That is, Gaussian probability distribution estimation generates a probability density function (PDF) in the form of a smooth curve in real space, and categorical probability distribution estimation can estimate the probability mass for each category in discrete space. However, in the present invention, different techniques are not necessarily applied to each parameter having different characteristics, and the Gaussian probability distribution estimation technique may be used for each of the first parameter and the second parameter.

[0091] Next, the data processing unit (160) may be configured to perform the role of generating (or configuring, constructing, etc.) at least one initial sample or candidate evaluation point by performing sampling from the estimated prior distribution for each of a plurality of parameters having different characteristics.

[0092] For example, the data processing unit (160) may generate multiple initial samples based on the prior distribution of each parameter estimated for each of the multiple parameters. Alternatively, the data processing unit (160) may extract initial samples from individual prior distributions estimated by the probability distribution estimation unit (150) using different probability distribution estimation techniques according to the data type and / or distribution characteristics of each of the multiple parameters. In this case, the data processing unit (160) may sample multiple initial samples from each of the first prior distribution estimated by performing Gaussian kernel density estimation (Gaussian KDE) on the first parameter and the second prior distribution estimated by performing categorical probability distribution estimation on the second parameter, and construct an initial sample set to be used for optimization using the multiple initial samples.

[0093] Next, the artificial intelligence model (170) can be trained based on a prior distribution estimated for multiple parameters having different characteristics. For example, the data processing unit (160) can construct an initial sample set using multiple initial samples, and the artificial intelligence model (170) can be trained using the initial sample set. In the present invention, the artificial intelligence model (170) may also be named a “probability model,” a “surrogate model,” a “proxy model,” or a “prediction model,” etc.

[0094] An artificial intelligence model (170) can be trained based on at least one of the result data generated through a first optimization algorithm (e.g., Evolutionary Algorithm (EA)) and a prior distribution for each of a plurality of parameters having different characteristics in order to perform optimization on the parameters to be optimized.

[0095] Meanwhile, as discussed above, combinatorial optimization problems have a structural limitation in that existing optimization methods based on continuous variables are difficult to apply directly, as the search target often consists of discrete variables. To overcome the above-mentioned limitation, the present invention,

[0096] Various candidate solutions are searched using a first optimization algorithm (e.g., an evolutionary algorithm), and result data is generated from the search results. By utilizing this as training data for an artificial intelligence model (170), a prediction model that reflects complex interactions between parameters can be constructed. The artificial intelligence model trained in this way can estimate the prediction performance and variance of each parameter combination, and by providing this as input to a second optimization algorithm, a combinatorial optimization framework can be implemented that can simultaneously improve search efficiency and the quality of the solution. That is, the present invention provides a technical effect that improves the problem of reduced search efficiency and local convergence in existing combinatorial optimization algorithms through a cyclic structure of “search-learn-research,” and simultaneously enables the stable generation of a global optimal solution even in a complex high-dimensional search space. Furthermore, since the artificial intelligence model is directly involved in the control of the search space and the selection process of candidate solutions, the present invention can realize the application of artificial intelligence as a technical means that goes beyond simple mathematical calculation procedures.

[0097] In the present invention, the artificial intelligence model (170) may also be named a “probability model,” a “surrogate model,” a “proxy model,” or a “prediction model.” More specific details regarding the learning of the artificial intelligence model (170) and the inference process of the learned artificial intelligence model (170) will be described later.

[0098] Next, the control unit (180) can perform the role of controlling the overall operation of the combination optimization system (100) related to the present invention. The control unit (180) can process signals, data, information, etc. that are input or output through the components of the combination optimization system (100) described above, or perform a series of data processing to provide or process appropriate information and functions to the user. The control unit (180) can be physically implemented by the processor described above.

[0099] Meanwhile, the present invention aims to provide a combinatorial optimization method and system capable of solving optimization problems and improving optimization performance. More specifically, the present invention aims to provide a combinatorial optimization method and system capable of extensively exploring an initial solution space and performing efficient and accurate parameter optimization. Below, we will examine the combinatorial optimization method according to the present invention in more detail.

[0100] In the present invention, input values ​​related to parameters to be optimized are specified (S210), and a process of searching for a plurality of candidate solutions generated according to the input values ​​can be performed using a first optimization algorithm that is pre-set (S220, see FIG. 2). Furthermore, in the present invention, result data is generated based on the results of searching for a plurality of candidate solutions (S230), and a process of training an artificial intelligence model (170) using the generated result data can be performed (S240, see FIG. 2).

[0101] Here, the parameters to be optimized may be included in a combination optimization system (100) or a target system for solving combination optimization problems (e.g., a simulator, a simulation system). As an example, the simulator may perform the role of determining (or predicting) whether a product to be manufactured is normal (21) or defective (22) (e.g., normal (21) or defective (22)). In the present invention, the term “parameter” may be used interchangeably with “variable”.

[0102] In this case, the simulator may be at least one component included in the combination optimization system (100) itself. Alternatively, the combination optimization system (100) may also be named a “simulation system” or “simulator” that performs combination optimization. The present invention is not limited to any one of these.

[0103] That is, the parameters to be optimized may refer to input variables (parameters) used in the combination optimization system (100, or simulator). For example, when the combination optimization system (100) receives various parameters (e.g., temperature, pressure, speed, lighting conditions, etc.) for process optimization, the control unit (180) may specify them as multiple parameters having different characteristics.

[0104] Additionally, the parameters to be optimized may include parameters received from a user terminal (10) in addition to parameters set (or defined, input) within the combination optimization system (100, or simulator). For example, when the control unit (180) receives input from the user terminal (10) regarding a first parameter (continuous variable) and a second parameter (discrete variable) to be optimized, the first parameter and the second parameter may be identified as a plurality of parameters having different characteristics.

[0105] However, in addition to the cases mentioned above, there may be various ways in which the parameters to be optimized are specified, and the present invention is not limited to any one of these.

[0106] Meanwhile, the first optimization algorithm in the present invention may include an “Evolutionary Algorithm (EA),” which is a metaheuristic optimization technique inspired by the process of natural selection. The Evolutionary Algorithm (EA) performs global optimization through population-based search. More specifically, the Evolutionary Algorithm is used to solve global optimization problems in various domains, and through population-based search, this Evolutionary Algorithm can generate various initial solutions even for complex problems (or instances).

[0107] As illustrated in FIG. 7(a), the control unit (180) can perform initialization for a first optimization algorithm (e.g., an evolutionary algorithm (EA)). For example, during the initialization process, input values ​​related to the parameters to be optimized may be specified, and a process of generating a plurality of candidate solutions (or a plurality of different candidate solutions) according to the specified input values ​​may be carried out. Here, the input values ​​related to the parameters to be optimized may be specified based on user input entered through the user terminal (10). Alternatively, the input values ​​related to the parameters to be optimized may be input values ​​internally set in the combinational optimization system (100). However, in the present invention, there may be various methods (or ways) for specifying input values ​​related to the parameters to be optimized in addition to those mentioned, and the present invention is not limited to any one of them.

[0108] More specifically, the input value related to the parameter to be optimized may include at least one of the number of candidate solutions containing the parameter to be optimized, the number of parameters to be optimized included in the candidate solutions, and the search range of the parameter to be optimized.

[0109] The number of candidate solutions (solutions) containing the optimization target parameters may refer to the total number of candidate solutions (or individuals, parameter combinations, etc.) to be generated in the initial generation. In the present invention, this may also be referred to as the “population size.” For example, if the population size included in the input value is “”, the control unit (180) may randomly generate a plurality (e.g., 50) of different candidate solutions (parameter combinations) within a search space defined based on the optimization target parameters. In the first optimization algorithm, an individual refers to one candidate solution of the optimization problem and may be expressed in the form of a vector containing the values ​​of multiple parameters (or variables) (e.g., [x1, x2, x3, ...]). That is, an individual may refer to one candidate solution (or one point) within the search space.

[0110] In addition, the number of optimization target parameters included in a candidate solution may refer to the number of parameters included (or possessed) by a single candidate solution. In the present invention, this may also be referred to as the “dimension number.” For example, if the optimization target parameters include lighting intensity, camera angle, and resolution, the dimension number may be “”. Each parameter included in a candidate solution may have its initial value randomly set within the search range. That is, the number of optimization target parameters included in a candidate solution may refer to the number of optimization target parameters included in the candidate solution (parameter combination) generated according to the input value.

[0111] Furthermore, the search range of the parameters to be optimized may refer to the minimum to maximum range of possible values ​​for each parameter (e.g., lighting intensity: 0 to 100, camera angle: 0° to 90°, image processing speed: 10fps to 120fps, resolution: 720p to 4K, etc.). In the present invention, when initializing the first optimization algorithm, multiple candidate solutions may be constructed by generating values ​​randomly within the search range of each parameter. This search range may serve to define the boundary of the entire parameter space (or search space). That is, the search space may refer to the entire area formed by combinations of possible values ​​of all parameters, and the search range may refer to the minimum to maximum value range for each parameter that defines the search space. As an example, if the number of dimensions is N, there are N search ranges (e.g., lighting intensity: 0 to 100, camera angle: 0° to 90°, image processing speed: 10fps to 120fps, resolution: 720p to 4K, etc.), and the entire area multiplied by the range of N can be the search space.

[0112] In this regard, the control unit (180) can generate multiple candidate solutions based on a specified input value (S301). The control unit can generate multiple candidate solutions by performing random sampling based on the input value (e.g., number of generated, number of dimensions, search range, etc.). For example, if the population size included in the input value is “” and the number of dimensions is “”, the control unit (180) can generate multiple (e.g., 50) different candidate solutions (e.g., [lighting intensity, camera angle, resolution]) during the first optimization algorithm initialization process. At this time, the control unit (180) can generate multiple candidate solutions based on the search range of each of the optimization target parameters (or multiple parameters having different characteristics) included in the input value.

[0113] Meanwhile, the control unit (180) can perform a search for a plurality of candidate solutions generated according to the input value using the first optimization algorithm.

[0114] In the present invention, performing a search using a first optimization algorithm may include an iterative process of evaluating candidate solutions individually and evolving into better candidates. In the first optimization algorithm, this process is performed in units of generations (e.g., number of iterations), and all candidate solutions may be evaluated and updated (or refreshed) in each generation. Alternatively, it may mean a process of using the first optimization algorithm to globally search an initial solution space defined based on a plurality of candidate solutions to select (or identify, select, etc.) an optimal candidate solution (or promising solution), storing the optimal candidate solution and the evaluation results for said optimal candidate solution (e.g., fitness values), and then providing them to a subsequent step (e.g., Bayesian optimization).

[0115] Specifically, the control unit (180) can repeatedly perform a search using the first optimization algorithm based on a preset number of iterations for the first optimization algorithm, and generate result data based on the search results repeatedly performed according to the preset number of iterations. Here, the preset number of iterations may be set based on user input entered through the user terminal (10). Alternatively, the preset number of iterations may be internally set and exist within the combination optimization system (100). However, it is obvious that such a number of iterations may be set in various ways by the administrator of the combination optimization system (100) or by the user utilizing the combination optimization system (100).

[0116] The control unit (180) can perform a search using the first optimization algorithm N times, corresponding to a preset number of repetitions N, and generate result data based on the search results performed N times. For example, the search using the first optimization algorithm may have “evaluate multiple candidate solutions (or evaluate updated candidate solutions) → select the optimal candidate solution based on the evaluation results → update multiple candidate solutions” as one cycle (or repetition, round, step, etc.), and in the next cycle, the candidate solutions updated in the previous cycle may be evaluated again. In this case, the candidate solutions updated in the previous cycle may be the subject of evaluation in the next cycle.

[0117] As illustrated in FIG. 7(b), the control unit (180) may perform a search for each of the multiple candidate solutions to select an optimal candidate solution (or a globally promising candidate solution) that satisfies a preset criterion from the multiple candidate solutions. Here, the initial population refers to the multiple candidate solutions, and the number of generations may refer to the number of iterations preset for the first optimization algorithm. In this search process, for each generation, the process of “evaluating each of the multiple candidate solutions → selecting the optimal candidate solution based on the evaluation results → updating the multiple candidate solutions” may proceed sequentially (or in order). That is, the initial population refers to the starting point (a set of various initial solutions) of the search using the first optimization algorithm, and the number of generations may refer to the number of iterations for improving and / or evolving the population.

[0118] In addition, the statement that a search is performed for each of the plurality of candidate solutions in the present invention can also be understood as performing a search for the initial solution space. The initial solution space may refer to a set of multiple candidate solutions generated through the first optimization algorithm initialization process described above.

[0119] The control unit (180) can perform a search for each of the multiple candidate solutions and generate result data based on the results of the search for each of the multiple candidate solutions.

[0120] First, as illustrated in FIG. 3, in the first iteration (or first cycle, first generation, first stage, first repetition, first search process, etc.) corresponding to the initial search process (or stage), the control unit (180) performs an evaluation for each of the plurality of candidate solutions and calculates an evaluation result for each of the plurality of candidate solutions (S303), and based on the evaluation result for each of the plurality of candidate solutions, can select (or specify, select, etc.) the optimal candidate solution among the plurality of candidate solutions that satisfies a preset criterion (S305). In this case, the result data may include at least one of the plurality of candidate solutions and the evaluation result for each of the plurality of candidate solutions.

[0121] The control unit (180) performs an evaluation of each of the multiple candidate solutions through a preset evaluation method (e.g., simulation) and, based on the evaluation results, can obtain (or collect) evaluation results for each of the multiple candidate solutions (e.g., actual evaluated objective (or target) function values ​​or evaluation scores or fitness values ​​for each of the multiple candidate solutions). Here, the evaluation results can also be understood as fitness values ​​calculated by inputting each of the multiple candidate solutions into a combination optimization system (100, or simulator) in the data processing unit (160). These fitness values ​​may include performance indicators for each of the multiple candidate solutions (e.g., accuracy, processing speed, quality score) indicating how well each of the multiple candidate solutions solves the problem.

[0122] In this case, the pre-established criteria for selecting the optimal candidate solution from multiple candidate solutions may be related to the evaluation results for each of the multiple candidate solutions. That is, the pre-established criteria may be set so that the candidate solution having the highest calculated fitness value among the multiple candidate solutions is selected as the optimal candidate solution. For example, the optimal candidate solution selected in the first search process in the present invention may also be named the “first optimal candidate solution.”

[0123] Furthermore, the control unit (180) can update at least some of the multiple candidate solutions using a preset operation method (S307, see FIG. 3). Here, the preset operation method may include a genetic operator. For example, the preset operation method may include a mutation operation. However, the operations included in the preset operation method are not necessarily limited to those mentioned, and may include various other operation methods (e.g., crossover, selection, etc.) in addition to those mentioned.

[0124] The control unit (180) can update all or at least some of a plurality of candidate solutions (or populations) (e.g., lower candidate solutions with low fitness values) through a mutation operation. Here, the process of updating candidate solutions through a mutation operation may include the process of generating new candidate solutions (or updated candidate solutions) by modifying the candidate solutions of each generation according to a certain rule, and configuring them as a set of candidate solutions (or populations) to be used in the next generation. Alternatively, it may include the process of probabilistically modifying some of the candidate solutions to diversify existing candidate solutions and expand the search range for each generation. That is, the mutation operation may modify (or replace) candidate solutions to evolve into the next generation.

[0125] In one embodiment, the control unit (180) may be a method of randomly changing some of the values ​​of the parameters constituting the solution through a variation operation (e.g., randomly adjusting the parameter corresponding to the lighting intensity within a range of ±5%).

[0126] In another embodiment, the control unit (180) may generate a new candidate solution by exchanging parameters constituting each of a plurality (e.g., 2) of candidate solutions through cross-operation (e.g., camera angle is collected from candidate solution A, resolution value is collected from candidate solution B and combined).

[0127] In this way, the present invention generates new candidate solutions through mutation operations, thereby maintaining data diversity and expanding the search range, and can enhance the globality of the optimization search.

[0128] Next, after the first search step is completed, a second search step corresponding to the next search step after the first can be performed. As illustrated in FIG. 3, in the second step (or second cycle, second generation, second stage, second iteration, second search process, etc.), the control unit (180) performs an evaluation of an updated candidate solution including a first optimal candidate solution to produce an evaluation result for the updated candidate solution (S309), and based on the evaluation result for the updated candidate solution, selects an optimal candidate solution among the updated candidate solutions that satisfies a preset criterion (S311). In this case, the result data may further include at least one of the updated candidate solution and the evaluation result for the updated candidate solution.

[0129] The control unit (180) performs an evaluation of each updated candidate solution through a preset evaluation method (e.g., simulation) and, based on the evaluation results, can obtain (or collect) evaluation results for each updated candidate solution (e.g., actual evaluated objective (or goal) function values ​​or evaluation scores or fitness values ​​for each updated candidate solution). Here, the evaluation results can also be understood as fitness values ​​calculated by inputting each updated candidate solution into a combination optimization system (100, or simulator) in the data processing unit (160). These fitness values ​​may include performance indicators (e.g., accuracy, processing speed, quality score) for each updated candidate solution that indicate how well each updated candidate solution solves the problem.

[0130] In this case, the pre-set criteria for selecting the optimal candidate solution from the updated candidate solutions may be related to the evaluation results for each of the updated candidate solutions. That is, the pre-set criteria may be configured so that the candidate solution having the highest calculated fitness value among the updated candidate solutions is selected as the optimal candidate solution. For example, the optimal candidate solution selected in the second search process in the present invention may also be referred to as the “second optimal candidate solution.”

[0131] And, the control unit (180) can update all or at least part of the updated candidate solutions including the second optimal candidate solution through mutation operations. Since the subsequent process is identical to what was described above, it will be omitted to avoid duplication of explanation. That is, the search process using the first algorithm examined above can be performed for a predetermined number of iterations.

[0132] As seen above, the search process using the first optimization algorithm may be a process of searching (or selecting) at least one optimal candidate solution for each iteration based on a preset number of iterations for the first optimization algorithm. In this case, the optimal candidate solution searched for each iteration may be included in the result data and used for training the artificial intelligence model (170). That is, the result data may include at least one of the optimal candidate solution searched for each iteration, the evaluation result for the optimal candidate solution, and the search range of the optimal candidate solution.

[0133] Here, selecting an optimal candidate solution (or optimal object) for each set number of iterations (or each generation) may mean that the task of selecting a candidate solution with the highest fitness value among the current candidate solutions is repeated for each iteration. The control unit (180) may store the optimal candidate solution selected for each iteration in a specified storage (e.g., storage unit (140, or memory)). In this case, the optimal candidate solution selected for each iteration may be maintained through an elitism method or used as a reference value in the mutation operation process. This can be understood as maintaining the optimal candidate solution of the previous generation in the next generation so as not to lose promising solutions and to gradually improve the solution during the search process.

[0134] In one embodiment, referring to FIG. 7(b), the optimal entity (i.e., optimal candidate solution) finally output through a search using the first optimization algorithm may correspond to any one of the optimal candidate solutions having the highest fitness value among the optimal candidate solutions selected for each iteration after all search processes according to a preset number of iterations have ended. That is, the optimal candidate solution finally returned through the first optimization algorithm may correspond to the best candidate solution (e.g., Global Best) among the optimal candidate solutions selected for each iteration. Such an optimal candidate solution may refer to the solution (or parameter combination, parameter set, etc.) having the highest globally highest fitness value finally obtained (or output) during the search process using the first optimization algorithm. Let us assume that there exists a first optimal candidate solution, a second optimal candidate solution, a third optimal candidate solution, and an Nth optimal candidate solution. The control unit (180) can use the optimal candidate solution having the highest fitness value among the first optimal candidate solution, the second optimal candidate solution, the third optimal candidate solution, and the Nth optimal candidate solution, based on the fitness value for each of the first optimal candidate solution, the second optimal candidate solution, the third optimal candidate solution, and the Nth optimal candidate solution, for the learning and Bayesian optimization process of the artificial intelligence model (170).

[0135] Meanwhile, as illustrated in FIG. 4, the present invention can generate result data based on the results of a search performed using a first optimization algorithm and store the generated result data in a specified storage (S401).

[0136] Here, the result data may include parameters included in each of all candidate solutions searched for each iteration after performing a search using the first optimization algorithm, and evaluation results for said parameters. For example, the result data may include the values ​​of each optimization target parameter included in all multiple candidate solutions searched through the first optimization algorithm, and the fitness values ​​of each said optimization target parameter. Alternatively, it may include data comprising a pair of optimization target parameters included in each of the multiple candidate solutions and the values ​​of said optimization target parameters. That is, the result data may include performance records obtained while performing the optimization process over multiple generations.

[0137] In this case, the parameter values ​​can refer to the setting values ​​associated with each parameter (e.g., lighting intensity, camera angle, resolution, etc.). Additionally, the fitness value can refer to a performance indicator representing how well the corresponding candidate solution (combination of parameters) solves the problem.

[0138] Additionally, the control unit (180) may use the optimal candidate solution and the fitness value for the optimal candidate solution discussed above to perform optimization on the parameter to be optimized. In this case, during the optimization process for the parameter to be optimized, the control unit (180) may use the optimal candidate solution and the fitness value for the optimal candidate solution as initial data points. Through this, in the optimization process according to the present invention, the search for optimal parameters through Bayesian optimization can be started from a promising region, thereby minimizing the cold start problem and allowing for faster entry into the local optimization stage. That is, in the Bayesian optimization process described below, the most promising solution found through the first optimization algorithm can be used as a starting point to precisely search the surrounding region.

[0139] Meanwhile, the control unit (180) can estimate a prior distribution for the optimization target parameter using a preset probability distribution estimation technique. The optimization target parameter may include multiple parameters having different characteristics. In this case, the control unit (180) can estimate a prior distribution for each of the multiple parameters using a preset probability distribution estimation technique.

[0140] The control unit (180) can estimate a prior distribution for each of a plurality of parameters having different characteristics by using the probability distribution estimation unit (150). In the present invention, the prior distribution may also be referred to as a “probability distribution” or “prior knowledge.”

[0141] The probability distribution estimation unit (150) can estimate individual probability density functions (PDFs) by applying a probability distribution estimation technique set for each of the multiple parameters in a mixed parameter (or variable) environment. A probability density function is a function that indicates the probability that a value of a continuous random variable exists in a specific interval, and can be understood as a mathematical expression of a probability distribution. The probability density function (p(x)) has a value greater than or equal to 0 for all possible values, and the integral value over the entire domain can be 1. For any interval ([a, b]), the probability that a random variable exists in that interval is given by the value obtained by integrating the probability density function over that interval. In the present invention, the probability density function of each parameter is estimated through a set probability distribution estimation technique, and this can be utilized as a functional expression of the prior distribution.

[0142] The probability distribution estimation unit (150) can estimate a prior distribution for each of a plurality of parameters having specific different characteristics using a pre-set probability distribution estimation technique.

[0143] Here, the prior distribution may refer to prior knowledge regarding the probability distribution of a parameter under analysis prior to the observed data in Bayesian statistics or Bayesian optimization. Alternatively, the prior distribution may refer to the probability distribution established for each parameter prior to the observation data in Bayesian statistics or Bayesian optimization. For example, in the present invention, the prior distribution may include probability density functions estimated individually (independently) by applying a pre-established probability distribution technique to each of the first and second parameters having different characteristics. This prior distribution is utilized for initial sampling and training of the prediction model by reflecting the characteristics and range of each variable, and can be updated in real-time or incrementally during the optimization process.

[0144] Specifically, the probability distribution estimation unit (150) can estimate a prior distribution for each of the plurality of parameters by applying a pre-set probability distribution estimation technique to each of the plurality of parameters. Here, the plurality of parameters may include at least one of a first parameter (or a first type of parameter) having a first characteristic and a second parameter (or a second type of parameter) having a second characteristic. The parameter having the first characteristic may include a first parameter having the characteristics of a continuous variable, and the parameter having the second characteristic may include a second parameter having the characteristics of a discrete variable.

[0145] The probability distribution estimation unit (150) can estimate a prior distribution for each of the first parameter and the second parameter by applying a pre-set probability distribution estimation technique to each of the first parameter and the second parameter having a first characteristic. In this case, the probability distribution estimation unit (150) can estimate a prior distribution for each of the first parameter and the second parameter by reflecting the characteristics of each of the first parameter and the second parameter, applying a pre-set first probability distribution estimation technique to the first parameter and applying a pre-set second probability distribution estimation technique to the second parameter. That is, the first parameter can be modeled through the first probability distribution estimation technique, and the second parameter can be modeled through the second probability distribution estimation technique.

[0146] In one embodiment, let us assume that among a plurality of parameters, a first parameter and a third parameter exist among the parameters having a first characteristic, and a second parameter and a fourth parameter exist among the parameters having a second characteristic. The probability distribution estimation unit (150) can estimate a prior distribution for each of the first parameter and the third parameter by applying a first probability distribution estimation technique to each of the first parameter and the third parameter having the first characteristic. Additionally, the probability distribution estimation unit (150) can estimate a prior distribution for each of the second parameter and the fourth parameter by applying a second probability distribution estimation technique to each of the second parameter and the fourth parameter having the second characteristic.

[0147] That is, the probability distribution estimation unit (150) can estimate the prior distribution for each parameter by performing kernel density estimation (KDE) for each parameter having different characteristics (see FIG. 5 (a)). This prior distribution estimation process can be represented as FIG. 5 (b). Here, the prior distribution (or probability density function) estimated for each parameter can be represented as FIG. 5 (c), and the sample size, bandwidth, and kernel function (e.g., first probability distribution estimation technique or second probability distribution estimation technique, etc.) can be represented as FIG. 5 (d) to (f), respectively.

[0148] Meanwhile, the process of constructing an initial sample set based on the prior distribution estimated for multiple parameters can be carried out.

[0149] Once the prior distribution estimation for individual parameters is complete, the control unit (180) can use the data processing unit (160) to construct an initial sample set to be used for training the artificial intelligence model (170).

[0150] In this case, the prior distribution estimated for each of the multiple parameters can be used as a weight to directly determine the selection probability of the corresponding parameter value during the sampling process. More specifically, for a specific interval (or region, part, zone, etc.) where the probability value (or probability density value, probability density function value, prior distribution value, etc.) is high (or large) in the probability density function (i.e., prior distribution) estimated (or calculated, obtained, etc.) through kernel density estimation (KDE), the number of samplings (or frequency) in said specific interval may be set to increase relatively based on (or proportionally to) that probability value. For example, among the prior distributions estimated for each of the multiple parameters, the interval with a large probability density value may have a high weight in the sampling probability.

[0151] In this way, by performing random sampling or importance sampling based on set weights, intervals with high prior distributions can have more samples selected relative to the same sample size. Conversely, intervals with low prior distributions are selected less frequently, which can prevent inefficient searching within the entire search space.

[0152] This sampling method is performed independently for each of the multiple parameters, and the sampling results (multiple initial samples) performed independently for each of the multiple parameters can be combined to form an initial sample set.

[0153] For example, an interval with a prior distribution value of 0.8 can be selected with a frequency approximately four times greater than an interval with a prior distribution value of 0.2. This allows for the concentration of search resources on intervals with high probability, thereby inducing faster convergence. As a result, the present invention reduces the number of unnecessary evaluations and enables the attainment of a higher-quality optimal solution within a given evaluation budget. In other words, the present invention provides the effect of reducing unnecessary evaluations and improving the speed of optimization convergence by concentrating search resources on intervals with high validity within the entire search space. Here, the search space may refer to the set of all candidate solutions (e.g., possible states, parameter combinations, paths, etc.) that can be explored to solve a given specific problem (or instance). This is expressed as all possible combinations of variables, parameters, states, or input values ​​that constitute the problem, and each point (or point) within that space may represent a single candidate solution. That is, the search space may refer to the set of all possible combinations generated according to predefined variables and the range of values ​​of those variables to derive an optimal solution for the problem.

[0154] In this regard, the data processing unit (160) can generate initial samples based on the estimated prior distribution for each of the plurality of parameters. More specifically, as illustrated in FIG. 3, the data processing unit (160) can sample the value of each of the plurality of parameters (or the value for each of the plurality of parameters (or parameter value), the value corresponding to each of the plurality of parameters (or parameter value)) from the prior distribution for each of the plurality of parameters, and generate a plurality of initial samples using the sampled value of each of the plurality of parameters.

[0155] In one embodiment, let us assume that there exists a first prior distribution (e.g., a Gaussian KDE with a mean of 75°C and a standard deviation of 5°C) estimated from a first parameter (e.g., temperature) having a first characteristic and a third prior distribution (e.g., a Gaussian KDE with a mean of 1.0 bar and a standard deviation of 0.1 bar) estimated from a third parameter (e.g., pressure), and there exists a second prior distribution (e.g., {A1, A2, B1, B2}) estimated from a second parameter (e.g., part specification code) having a second characteristic and a fourth prior distribution (e.g., {Step 1: 0.4, Step 2: 0.35, Step 3: 0.25}) estimated from a fourth parameter (e.g., process step). The data processing unit (160) can perform sampling (e.g., random sampling) from estimated prior distributions for each of a plurality of parameters having different characteristics (e.g., first parameter, second parameter, third parameter, fourth parameter, etc.).

[0156] In this case, sampling is performed independently for the prior distributions estimated for multiple parameters a predetermined number of times (N times), and the results of such sampling are combined to generate N initial samples. That is, sampling is performed the same number of times N for the prior distribution of each parameter, and N multidimensional initial samples can be generated by combining identical indices. Here, the meaning of performing N times independently implies that the process of extracting initial samples from the prior distribution of each parameter is repeated N times through different random number extraction processes without interference (i.e., N samples are repeatedly extracted randomly from the same distribution). This means that when sampling the first parameter, the values ​​of the second, third, and fourth parameters are not affected, and sampling for each parameter can be performed from independent probability distributions (prior distributions). In addition, when N=3, three values ​​are sampled from the prior distribution of each parameter (e.g., three values ​​from the prior distribution of the first parameter and five values ​​from the prior distribution of the second parameter), and each sample can be newly randomly based on extraction regardless of the previously extracted value.

[0157] In one embodiment, the data processing unit (160) may perform a first sampling of the estimated prior distribution for each of a plurality of parameters (e.g., a first parameter, a second parameter, a third parameter, a fourth parameter, etc.) to obtain the values ​​of each of the plurality of parameters corresponding to the first sampling result (e.g., 72.8°C, A1, 1.05 bar, 2nd step, etc.). In this case, the data processing unit (160) may combine (or combine) the values ​​of each of the plurality of parameters obtained as the first sampling result to construct (or generate) a first initial sample (e.g., [72.8°C, A1, 1.05 bar, 2nd step]).

[0158] In another embodiment, the data processing unit (160) may perform a second sampling of the estimated prior distribution for each of the plurality of parameters to obtain the values ​​of each of the plurality of parameters corresponding to the second sampling result (e.g., 76.3°C, B2, 0.98 bar, Step 1). In this case, the data processing unit (160) may combine (or combine) the values ​​of each of the plurality of parameters obtained as the second sampling result to form a second initial sample (e.g., [76.3°C, B2, 0.98 bar, Step 1]).

[0159] In another embodiment, the data processing unit (160) may perform a third sampling of the estimated prior distribution for each of the plurality of parameters to obtain the values ​​of each of the plurality of parameters corresponding to the third sampling result (e.g., 74.1°C, A2, 1.02 bar, step 1, etc.). In this case, the data processing unit (160) may combine (or combine) the values ​​of each of the plurality of parameters obtained as the third sampling result to form a third initial sample (e.g., [74.1°C, A2, 1.02 bar, step 1]).

[0160] Additionally, the data processing unit (160) can weight the sampling probability based on the probability density function value in the prior distribution for each of the plurality of parameters (i.e., adjust or set the weight of the sampling probability). The data processing unit (160) can specify a specific interval with high probability density according to (or based on) the prior distribution estimated for each of the plurality of parameters, and perform sampling so that the probability of a sample being selected increases in the interval specified as having high probability density (or so that the frequency of sample selection increases, so that the sample is preferentially selected).

[0161] In one embodiment, the data processing unit (160) can identify a specific interval with high probability density based on a first prior distribution (e.g., a Gaussian KDE with a mean of 75°C and a standard deviation of 5°C) estimated from a first parameter (e.g., temperature). In the case of the first parameter, based on the fact that a high probability density is formed in the average 75°C interval (e.g., the 70-80°C interval), the data processing unit (160) can perform N samplings by adjusting the weights so that the sampling probability in the specific interval (e.g., the 70-80°C interval) increases (or becomes higher).

[0162] Alternatively, the data processing unit (160) may specify a section with low probability density according to (or based on) the prior distribution estimated for each of the multiple parameters, and perform sampling so that the probability of a sample being selected in the section specified as having low probability density is reduced (or lowered).

[0163] In one embodiment, let us assume that the Nth prior distribution of the Nth parameter (e.g., speed) is estimated to have a mean of 1.5 m / s and a standard deviation of 0.2 m / s. The data processing unit (160) can identify a high probability density section and a low probability density section, respectively, based on the Nth prior distribution of the Nth parameter (e.g., speed). In this case, the 1.4 to 1.6 m / s section has a high probability density, but the 0.8 m / s or lower or 2.2 m / s or higher section may have a low probability density. The data processing unit (160) can identify the section having a low probability density and perform sampling so that the sampling probability in the identified section is reduced (or lowered). At this time, when N is set to 5, speed values ​​of 0.8 m / s or less and 2.2 m / s or more can be selected with equal probability when uniformly sampling the entire distribution, but the data processing unit (160) can set the sampling probability in the low probability density range to be close to 0 so that most of the initial samples sampled from the Nth parameter are intensively sampled in the high probability density range (e.g., 1.4 to 1.6 m / s).

[0164] Accordingly, regions with a relatively low probability of occurrence within the entire search space are naturally excluded, and samples are generated by concentrating them in regions with a high probability of occurrence. Compared to the case where initial samples are randomly and uniformly distributed, this improves optimization efficiency and reduces unnecessary evaluations by allowing promising regions to be explored with higher frequency. This approach contributes to increasing search efficiency in the subsequent Bayesian optimization process by constructing initial samples that reflect the prior distribution characteristics of each parameter. In other words, since the initial samples generated through this method are structured to reflect the prior distribution characteristics of each parameter while concentrating in regions with high probability density, they improve optimization efficiency during the subsequent Bayesian optimization process and enable the derivation of high-performance candidate solutions with fewer iterations.

[0165] In this case, the process of generating multiple initial samples may be a process of generating multiple initial samples including different parameter combinations (e.g., a first parameter combination, a second parameter combination, a third parameter combination, etc.) by using the values ​​of each of the multiple parameters sampled from the prior distribution for each of the multiple parameters. The initial samples may include parameter combinations sampled from the prior distribution estimated for each of the multiple parameters. That is, in the present invention, after estimating the prior distribution of individual parameters, the combination of parameters sampled from the prior distributions may be named an “initial sample.”

[0166] Alternatively, the process of generating (or constructing) a plurality of initial samples in the present invention may also be understood as a process of constructing different parameter vectors (e.g., [x1, x2, ... xn]) using values ​​sampled from a prior distribution for each of the plurality of parameters (e.g., values ​​of each of the plurality of parameters). In the present invention, the initial samples may also be referred to as “initial data,” “initial input sample,” “initial input data,” “initial parameter combination,” “initial sample vector,” “sample vector,” “training sample,” “training sample vector,” “input sample,” “input sample vector,” “input vector,” “initial sampling data (sampled data),” “initial parameter set,” “initial set,” “initial parameter pair,” or “training data,” etc.

[0167] In addition, the data processing unit (160) may construct (or generate, build, etc.) an initial sample set using a plurality of initial samples to train a prediction model. For example, the data processing unit (160) may construct an initial sample set including at least one of a first initial sample (e.g., [72.8°C, A1, 1.05 bar, Step 2]), a second initial sample (e.g., [76.3°C, B2, 0.98 bar, Step 1]), a third initial sample (e.g., [74.1°C, A2, 1.02 bar, Step 1]), and a Nth initial sample.

[0168] In this case, during the process of constructing the initial sample set, the Bayesian optimization system (100) may perform an evaluation on each of the multiple initial samples sampled from the prior distribution for each of the multiple parameters. The Bayesian optimization system (100) may perform an evaluation on each of the multiple initial samples through a pre-set evaluation method (e.g., simulation), and based on the evaluation results, obtain (or collect) the evaluation results for each of the multiple initial samples (e.g., the objective (or target) function value or evaluation score (or score) actually evaluated for the parameter combination included in each of the multiple initial samples). Here, the objective function value evaluated for the initial samples may also be understood as a performance indicator (or performance value) calculated by inputting each initial sample into the Bayesian optimization system (100, or simulator) in the data processing unit (160). As an example, the form of the value may be numeric (e.g., score, time cost, etc.), and may be a single value or, in the case of multi-objective optimization, a vector form.

[0169] Specifically, the Bayesian optimization system (100) can perform a simulation for each of a plurality of initial samples containing different parameter combinations to obtain an evaluation result for each of the plurality of initial samples.

[0170] For example, a Bayesian optimization system (100) can perform a simulation on a first parameter combination (e.g., [72.8°C, A1, 1.05 bar, 2 steps]) included in a first initial sample to obtain an evaluation result (e.g., 92.5) for the first parameter combination.

[0171] As another example, the Bayesian optimization system (100) can perform a simulation on a second parameter combination (e.g., [76.3°C, B2, 0.98 bar, step 1]) included in a second initial sample to obtain an evaluation result (e.g., 88.3) for the second parameter combination.

[0172] As another example, the Bayesian optimization system (100) can perform a simulation on a third parameter combination (e.g., [74.1°C, A2, 1.02 bar, 1 step]) included in a third initial sample to obtain an evaluation result (e.g., 95.7) for the third parameter combination.

[0173] Furthermore, the data processing unit (160) can construct an initial sample set using a plurality of initial samples and an evaluation result for each of the plurality of initial samples. For example, the data processing unit (160) can construct an initial sample set by having a plurality of initial samples and an evaluation result for each of the plurality of initial samples form a pair. In this case, the initial sample set may be configured to include at least one of a first initial sample-first evaluation result pair in which a first initial sample (X1) and a first evaluation result for said first initial sample form a pair, a second initial sample-second evaluation result pair in which a second initial sample (X2) and a second evaluation result for said second initial sample form a pair, a third initial sample-third evaluation result pair in which a third initial sample (X3) and a third evaluation result for said third initial sample form a pair, and an Nth initial sample-Nth evaluation result pair in which an Nth initial sample (XN) and a Nth evaluation result for said Nth initial sample form a pair (e.g., D = {(X1, 92.5), (X2, 88.3), (X3, 95.7)..., (XN, ...)}). In the present invention, the initial sample set may also be named “initial sample vector set,” “training sample set,” “training sample vector set,” “input sample set,” “input sample vector set,” or “training data set,” etc.

[0174] Meanwhile, in the present invention, a process of performing optimization on the parameters to be optimized based on a learned artificial intelligence model and a pre-set second optimization algorithm may be carried out (S250, see FIG. 2).

[0175] As illustrated in FIG. 4, the control unit (180) can train an artificial intelligence model (170) using at least one of a prior distribution for the parameter to be optimized and result data stored in a specified repository in order to perform optimization for the parameter to be optimized (S403). More specifically, the control unit (180) can train an artificial intelligence model (170) using at least one of a prior distribution for each of a plurality of parameters and result data.

[0176] In this case, the prior distribution corresponding to the kernel density estimation result and the result data generated based on the execution of the first optimization algorithm may differ from each other. More specifically, kernel density estimation estimates the frequency of occurrence of each value of multiple parameters based on all data collected in the past, and can form a distribution in which the probability density is concentrated in the interval with the highest frequency. On the other hand, the result data does not use the simple frequency of past data, but rather performs the search through the evaluation of the fitness value (or goodness of fit) of candidate solutions sampled in the current generation. Therefore, the distribution of the result data generated through the first optimization algorithm does not depend on past data and can dynamically change around parameter combinations that perform well during the current search process. Accordingly, even when considering the same parameter space (search space), the distribution formed by KDE and the distribution derived by EA may exhibit different aspects, and in the present invention, the two data can be combined to improve optimization performance and perform efficient optimization.

[0177] In one embodiment, the control unit (180) may train an artificial intelligence model (170) by using at least a portion of the result data stored in a specified repository and an initial sample set configured by performing sampling from a prior distribution estimated for each of a plurality of parameters as training data. The types of models that may be included in the artificial intelligence model (170) in the present invention may vary. For example, the types of artificial intelligence models (170) may include at least one of a Gaussian Process (GP) model and / or a Random Forest model and / or a Bayesian Neural Network. However, this is merely an example, and the types of models included in the artificial intelligence model (170) are not necessarily limited thereto. Additionally, the artificial intelligence model (170) may include one or multiple models. If multiple models are included, the multiple models may be combined into an ensemble (e.g., combining a Gaussian Process model and a Random Forest model) and used in the optimization process.

[0178] As described above, the artificial intelligence model (170) may be composed of a single model or multiple models, and if multiple models are included, these models may be combined complementarily to form an ensemble architecture that corrects prediction deviations between models (preventing overfitting) and improves prediction reliability. For example, a Gaussian Process (GP) model performs probabilistic inference that can quantify the uncertainty of prediction, a Random Forest model efficiently learns non-linear relationships through multiple decision tree-based averaging, and a Bayesian Neural Network can model complex non-linear relationships by estimating the probabilistic distribution of neural network parameters.

[0179] The control unit (180) can determine the final prediction result by aggregating the prediction value and prediction variance calculated from each model, and in this process, the model weighting factor can be dynamically adjusted in proportion to the past prediction accuracy or reliability of each model. Accordingly, the artificial intelligence model (170) of the present invention functions as a technical means that goes beyond simply statistically estimating data and directly intervenes in the logical decision-making of the search algorithm to control the selection of the search path and candidate solution.

[0180] Additionally, the artificial intelligence model (170) can calculate an acquisition function by considering all of the input parameter vector (feature vector), the evaluation result (objective value) of the objective function, and the prediction variance during the learning process. The acquisition function can be defined in the form of, for example, Expected Improvement (EI), Upper Confidence Bound (UCB), etc., and is provided to the control unit (180) to maximize exploration efficiency by quantitatively evaluating the exploration value of each candidate solution in the exploration space.

[0181] The trained artificial intelligence model (170) is linked with a second optimization algorithm to perform priority re-search on a portion of the search space or to propose new sampling candidates, thereby improving the overall search efficiency of the system. Furthermore, newly acquired result data during the iterative cycle is additionally stored in a designated repository and used as retraining data for the artificial intelligence model (170), thereby implementing a self-evolving optimization structure in which the model's prediction accuracy and search stability are gradually improved.

[0182] In terms of hardware, the artificial intelligence model (170) can be implemented by a processor (CPU, GPU, or NPU) and program code executed in memory. The learning phase of the model is performed based on parallel computation to improve the processing speed of large-scale search data, and in the inference phase, a lightweight inference model is used to minimize system latency when performing real-time optimization.

[0183] This artificial intelligence model (170) can learn an approximation of the optimization target function (f(x)) in an environment where multiple parameters are mixed, by utilizing the prior distribution for each of the multiple parameters. More specifically, instead of directly calculating the actual optimization target function, the artificial intelligence model (170) can approximate the optimization target function by learning an initial sample set (pairs of initial samples and evaluation results for those samples). That is, learning an approximation of the optimization target function can also be understood as the artificial intelligence model (170) being trained to approximate multiple initial samples and evaluation results for each of the multiple initial samples. Alternatively, the artificial intelligence model (170) can be understood as being trained in a direction that approximates the optimization target function (i.e., the actual objective function). In other words, the artificial intelligence model (170) can perform the role of providing an approximation of the optimization target function by learning multiple initial samples included in the initial sample set and evaluation results for each of the multiple initial samples.

[0184] When the training of the artificial intelligence model (170) is completed, the control unit (180) can obtain the trained artificial intelligence model (170) based on the training. Then, the control unit (180) can perform optimization on the search space based on the trained artificial intelligence model (170) and a second optimization algorithm (e.g., Bayesian optimization).

[0185] As seen above, in the process S210 to S230, the initial solution space is widely searched to calculate optimal candidate solutions (i.e., promising parameter combinations) and fitness values ​​of the optimal candidate solutions. The search results using this first optimization algorithm can be used as input data for training an artificial intelligence model (170) used in Bayesian optimization.

[0186] In one embodiment, as illustrated in FIG. 7(c), during the Bayesian optimization process, the control unit (180) can train an artificial intelligence model (170) based on at least one of an optimal candidate solution (e.g., optimal object) obtained through a first optimization algorithm, a fitness value of the optimal candidate solution, and a search range of the optimal candidate solution. Subsequently, the control unit (180) can perform precise local search by repeating the process of proposing and verifying new parameter combinations throughout the entire search space. In this regard, optimizing parameters based on the trained artificial intelligence model (170) and the second optimization algorithm can be understood as performing additional precise optimization within the entire search space (parameter space) starting from the optimal candidate solution. That is, it can be understood as setting the parameters of the artificial intelligence model (170) based on result data generated using the first optimization algorithm, and thereby proposing new parameters within the search space and performing the optimization process.

[0187] As discussed above, the search space can be defined based on the parameters to be optimized. For example, the search space can be defined according to the input values ​​associated with the parameters to be optimized.

[0188] The control unit (180) can perform optimization on the optimization target parameter through at least one combination of parameters specified (or selected, decided, chosen, etc.) from the search space using a learned artificial intelligence model (170) and a specific function.

[0189] For example, a specific function may include an acquisition function. The control unit (180) may calculate a specific function value for the search space based on a learned artificial intelligence model (170) to specify at least one parameter combination from the search space. At this time, the specific function may serve to quantify how much search value each candidate point (candidate parameter combination) has. And, the control unit (180) may specify (or determine) a candidate point where the acquisition function value is maximized. A candidate point specified from this search space may also be understood as a “candidate parameter combination to be newly sampled.” That is, the specific function calculates a specific function value for all candidate points (or points considered during the optimization process) defined in the search space, and the point with the maximum value among them may become the parameter combination to be newly sampled. Here, a candidate point may refer to a specific parameter combination in a specific search space. Such a parameter combination may also be expressed as a “parameter vector.”

[0190] Specifically, the control unit (180) can generate (or extract, sample, etc.) a new parameter combination using a specific function. Here, generating a new parameter combination may mean selecting the next parameter combination to be evaluated within the search space based on the value calculated by the specific function. Alternatively, it may mean the process of calculating a candidate combination of the optimization target parameters to be evaluated next based on the artificial intelligence model (170) learned in the Bayesian optimization process. Alternatively, it may mean sampling a new parameter combination based on the specific function. Such a new parameter combination may mean a combination of the optimization target parameters (a single vector).

[0191] For example, the control unit (180) can use the learned artificial intelligence model (170) and the search function to evaluate how good the performance is when searching for each parameter combination (or each point, each candidate point, etc.) included in the search space. In this case, the control unit (180) can select the parameter combination with the highest search function value as the next parameter combination to be evaluated. At this time, selecting the parameter combination to be evaluated can also be understood as a process of sampling a new candidate parameter combination in the search space (i.e., sampling a parameter combination at the point where the search function value is maximum). Afterward, the control unit (180) evaluates the sampled parameter combination and can update the artificial intelligence model (170) by reflecting the evaluation result in the learned artificial intelligence model (170).

[0192] That is, the control unit (180) can calculate the predicted mean (performance expectation) and predicted variance (or uncertainty, standard deviation, etc.) for each of the candidate parameter combinations in the search space using the learned artificial intelligence model (170), and apply a specific function to each of the candidate parameter combinations to quantify the specific function value. Then, the control unit (180) can select the parameter combination in which the specific function value is maximized, perform an evaluation of the selected parameter combination, and then reflect the evaluation result of the selected parameter combination in the learned artificial intelligence model (170). That is, the control unit (180) can sample the parameter combination at the point where it is judged to be most promising with the specific function, and update the artificial intelligence model (170) by reflecting the evaluation result of the sampled parameter combination in the learned artificial intelligence model (170).

[0193] Furthermore, as illustrated in FIG. 4, the control unit (180) updates the learned artificial intelligence model (170) based on a combination of parameters specified from the search space (S405), and based on the update to the learned artificial intelligence model (170), can perform optimization on the parameters to be optimized (S407).

[0194] The control unit (180) can perform optimization on the parameters to be optimized based on the updated artificial intelligence model and the second optimization algorithm to produce an optimal solution for the parameters to be optimized. This optimal solution (i.e., the final optimized parameters) may refer to a combination of the parameters to be optimized. For example, the optimal solution may be configured to include a first parameter, a second parameter, a third parameter, and an Nth parameter included in the parameters to be optimized. In this case, the optimization process for the parameters to be optimized using the Bayesian optimization described above may be performed repeatedly until the optimized parameters (i.e., the optimal solution) are produced (or derived).

[0195] In another embodiment, when the control unit (180) acquires the learned artificial intelligence model (170), it may sample candidate evaluation points using the data processing unit (160). The data processing unit (160) may sample candidate evaluation points from the prior distribution for each of the plurality of parameters. The method of sampling candidate evaluation points in the present invention may be the same as the method of sampling the initial sample described above. Accordingly, to avoid duplication of explanation, the method of sampling candidate evaluation points will be described only briefly below.

[0196] The data processing unit (160) can sample candidate evaluation points based on the estimated prior distribution for each of the plurality of parameters. Alternatively, the data processing unit (160) can generate candidate evaluation points from the prior distribution for each of the plurality of parameters.

[0197] In one embodiment, the data processing unit (160) may perform a first sampling of the estimated prior distribution for each of the plurality of parameters (e.g., a first parameter (e.g., temperature), a second parameter (e.g., part specification code), a third parameter (e.g., pressure), a fourth parameter (e.g., process step), etc.) to obtain the values ​​of each of the plurality of parameters corresponding to the first sampling result (e.g., 73.8°C, A2, 1.05 bar, step 2, etc.). In this case, the data processing unit (160) may combine (or combine) the values ​​of each of the plurality of parameters obtained as the first sampling result to generate (or configure) a first candidate evaluation point (e.g., [73.8°C, A2, 1.05 bar, step 2]).

[0198] In another embodiment, the data processing unit (160) may perform a second sampling of the estimated prior distribution for each of the plurality of parameters to obtain the values ​​of each of the plurality of parameters corresponding to the second sampling result (e.g., 74.2°C, B1, 0.96 bar, 2nd stage). In this case, the data processing unit (160) may combine (or combine) the values ​​of each of the plurality of parameters obtained as the second sampling result to form a second candidate evaluation point (e.g., [74.2°C, B1, 0.96 bar, 2nd stage]).

[0199] In another embodiment, the data processing unit (160) may perform a third sampling of the estimated prior distribution for each of the plurality of parameters to obtain the values ​​of each of the plurality of parameters corresponding to the third sampling result (e.g., 74.4°C, A2, 1.04 bar, level 1, etc.). In this case, the data processing unit (160) may combine (or combine) the values ​​of each of the plurality of parameters obtained as the third sampling result to form a third candidate evaluation point (e.g., [74.4°C, A2, 1.04 bar, level 1]).

[0200] In this case, the process of generating candidate evaluation points may be a process of generating candidate evaluation points containing different parameter combinations by using the values ​​of each of the multiple parameters sampled from the prior distribution for each of the multiple parameters. In this case, the candidate evaluation points may include new multiple parameter combinations that are not included in the initial sample set used in the learning process of the artificial intelligence model (170). That is, the data processing unit (160) may generate candidate evaluation points containing different parameter combinations (or corresponding to different parameter combinations) by sampling parameter combinations that are not included in the initial sample set.

[0201] Candidate evaluation points may include combinations of parameters sampled from prior distributions estimated for each of the plurality of parameters. That is, in the present invention, after the artificial intelligence model (170) is trained, combinations of parameters sampled from prior distributions of individual parameters may be named “candidate evaluation points.”

[0202] Alternatively, the process of generating candidate evaluation points in the present invention may also be understood as a process of constructing different parameter vectors (e.g., [x1, x2, ... xn]) using values ​​sampled from a prior distribution for each of a plurality of parameters. In the present invention, a candidate evaluation point may also be referred to as a “candidate point,” “candidate evaluation data,” “candidate evaluation sample,” “candidate parameter combination,” “candidate sample vector,” “candidate sample,” “candidate input sample,” “candidate input sample vector,” “candidate input data,” “candidate vector,” “candidate input vector,” “candidate sampling data (sampled data),” “candidate parameter set,” “candidate evaluation set,” or “candidate parameter pair,” etc.

[0203] And, the control unit (180) can select at least one specific evaluation point among the candidate evaluation points to be evaluated using the learned artificial intelligence model (170) and a specific function.

[0204] The trained artificial intelligence model (170) can be configured to predict at least one of the predicted value and predicted variance for each of the candidate evaluation points.

[0205] Here, the predicted value (or function value) may be a predicted average value calculated by the learned artificial intelligence model (170) by approximating the value of the objective (or goal) function for a specific input (candidate evaluation point (x)). Instead of directly calculating or experimenting with the actual objective function, it can be understood as a result value estimated through an initial sample set and the artificial intelligence model (170) trained on said initial sample set. Such a predicted value can predict the performance of the candidate evaluation point in advance without actual function evaluation. The predicted value of such a learned prediction model can be represented as shown in (c) of FIG. 6.

[0206] Additionally, the prediction variance may be a variance representing the degree of uncertainty regarding the predicted value when the learned artificial intelligence model (170) predicts the objective function value at a specific input. The higher the prediction variance, the greater the uncertainty, as there is a lack of data in the corresponding input range or little experience with model training; the higher the exploration value, the greater the likelihood of acquiring new information (however, the model's predicted value may differ significantly from the actual value). On the other hand, the lower the prediction variance, the lower the uncertainty, as there is sufficient data in the corresponding input range and the model has learned well; the higher the utility value, the more advantageous it may be for fine-tuning around an already known optimal point (the model's predicted value is likely to approximate the actual value).

[0207] The control unit (180) can predict a predicted value for each of the candidate evaluation points using the learned artificial intelligence model (170). More specifically, the control unit (180) can input the candidate evaluation points into the learned artificial intelligence model (170) and obtain a predicted value for each of the candidate evaluation points from the learned artificial intelligence model (170). For example, when the learned artificial intelligence model (170) receives an input for a candidate evaluation point composed of multiple parameters having different characteristics, it can estimate a conditional probability distribution for the input using parameters (e.g., weights, kernel parameters, etc.) obtained through learning through an initial sample set. In this case, the learned artificial intelligence model (170) can finally output a predicted value (or predicted mean value) and a predicted variance for the candidate evaluation point. As seen above, the predicted value represents an approximation of the objective function at the candidate evaluation point, and the predicted variance can represent the degree of uncertainty of the approximation.

[0208] In one embodiment, the control unit (180) may input a first candidate evaluation point (e.g., [73.8°C, A2, 1.05 bar, 2nd step]) among candidate evaluation points into a trained artificial intelligence model (170). Then, the control unit (180) may obtain from the trained artificial intelligence model (170) at least one of a predicted value (e.g., 0.85) and a predicted variance (e.g., 0.05) predicted by a prediction model trained for the first candidate evaluation point.

[0209] In another embodiment, the control unit (180) may input a second candidate evaluation point (e.g., [74.2°C, B1, 0.96 bar, 2nd step]) among the candidate evaluation points into the learned artificial intelligence model (170). Then, the control unit (180) may obtain from the learned artificial intelligence model (170) at least one of the predicted value (e.g., 0.78) and the predicted variance (e.g., 0.10) predicted by the learned prediction model for the second candidate evaluation point.

[0210] In another embodiment, the control unit (180) may input a third candidate evaluation point (e.g., [74.4°C, A2, 1.04 bar, 1st step]) among the candidate evaluation points into the learned artificial intelligence model (170). Then, the control unit (180) may obtain from the learned artificial intelligence model (170) at least one of the predicted value (e.g., 0.90) and predicted variance (e.g., 0.02) predicted by the prediction model learned for the second candidate evaluation point.

[0211] Next, the control unit (180) can select a specific evaluation point using a predicted value and a specific function for each of the candidate evaluation points.

[0212] Specifically, the control unit (180) calculates a specific function value for each of the candidate evaluation points using at least one of the predicted value for each of the candidate evaluation points and the optimal value that serves as a criterion for selecting a specific evaluation point, and among the candidate evaluation points, selects (or specifies) at least one evaluation point where the calculated specific function value satisfies a preset criterion as a specific evaluation point. Here, the optimal value that serves as a criterion for selecting a specific evaluation point may refer to the function value of the point with the best target function value among all evaluation points obtained so far in the optimization process. This may be a value measured through experiments and / or simulations, or may correspond to at least one of the highest values ​​among the predicted values ​​predicted by the learned artificial intelligence model (170).

[0213] In the present invention, a specific function may further include an Expected Improvement (EI) function. The Expected Improvement (EI) function may be an acquisition function that probabilistically estimates the degree of expected performance improvement at a new candidate evaluation point compared to the optimal value observed so far during the Bayesian optimization process. This specific function may be represented as in FIG. 6 (a). Here, the expected value may be represented as in FIG. 6 (b), the predicted value of the artificial intelligence model (170) for each of the candidate evaluation points may be represented as in FIG. 6 (c), and the optimal value observed so far (or the optimal value serving as a criterion for selecting a specific evaluation point) may be represented as in FIG. 6 (d). At this time, FIG. 6 (e) may mean that it is reflected only when there is improvement.

[0214] The control unit (180) can calculate a specific function value for each of the multiple candidate evaluation points using the mathematical formula (i.e., a specific function) shown in (a) of FIG. 6.

[0215] In one embodiment, the control unit (180) can calculate a specific function value (e.g., 0.0542) for a first candidate evaluation point (e.g., [73.8°C, A2, 1.05 bar, 2nd step]) by using at least one of a predicted value (e.g., 0.85), a predicted variance (e.g., 0.05), and an optimal value (e.g., 0.80) for a first candidate evaluation point.

[0216] In another embodiment, the control unit (180) can calculate a specific function value (e.g., 0.0307) for a second candidate evaluation point (e.g., [74.2°C, B1, 0.96 bar, 2nd step]) using at least one of a predicted value (e.g., 0.78), a predicted variance (e.g., 0.10), and an optimal value (e.g., 0.80) for the second candidate evaluation point.

[0217] In another embodiment, the control unit (180) can calculate a specific function value (e.g., 0.1) for a third candidate evaluation point (e.g., [74.4°C, A2, 1.04 bar, 1st step]) using at least one of a predicted value (e.g., 0.90), a predicted variance (e.g., 0.02), and an optimal value (e.g., 0.80) for the third candidate evaluation point.

[0218] Furthermore, as a result of calculating a specific function value for each of the multiple candidate evaluation points, the control unit (180) may select at least one evaluation point among the multiple candidate evaluation points as a specific evaluation point, wherein the calculated specific function value satisfies a preset criterion. For example, the preset criterion (or preset selection criterion) may be related to a point among the multiple candidate evaluation points that maximizes the specific function value.

[0219] The control unit (180) may select a point that maximizes a specific function value among a plurality of candidate evaluation points as a specific evaluation point. Such a specific evaluation point may include a point that maximizes a specific function value. Here, selecting a point that maximizes a specific function value (or an acquired function value) may mean selecting a point with the highest improvement expectation, rather than simply a point with a high predicted value. The acquired function described above is a function that numerically expresses the expectation of being better than the current optimal value by considering both the predicted value and uncertainty predicted by the artificial intelligence model (170). In this case, a high improvement expectation may include at least one of the following: i) high prediction performance, ii) high possibility of improvement due to high uncertainty (prediction variance), or iii) both. That is, selecting a point that maximizes a specific function value can be understood as selecting the point with the most expected performance improvement in the next evaluation.

[0220] For example, the control unit (180) may select the third candidate evaluation point, which has the maximum (or highest) specific function value among the multiple candidate evaluation points, as the next specific evaluation point to proceed with evaluation, based on the specific function value for the first candidate evaluation point (e.g., 0.0542), the specific function value for the second candidate evaluation point (e.g., 0.0307), and the specific function value for the third candidate evaluation point (e.g., 0.1). However, the number of specific evaluation points selected is not necessarily limited to one. For example, if there are multiple candidate evaluation points with the same specific function value, multiple specific evaluation points may be selected.

[0221] As seen above, the artificial intelligence model (170) may refer to a model that predicts the value of an objective function by performing learning based on initial data and an estimated prior distribution, instead of directly evaluating the actual objective function during the optimization process. This artificial intelligence model (170) learns the interaction between multiple parameters having different characteristics and selects the next optimization evaluation point. That is, the artificial intelligence model (170) can reduce the total number of evaluations and improve optimization efficiency by using learned interaction information to select the next optimization evaluation point through an acquisition function (e.g., Expected Improvement (EI) function). Below, we will examine the process of updating the artificial intelligence model (170) based on the selected evaluation point.

[0222] Meanwhile, in the present invention, an evaluation is performed on a selected specific evaluation point, and a learned prediction model can be updated based on the evaluation results for the specific evaluation point.

[0223] As seen above, based on the learned artificial intelligence model (170) and a specific function, candidate evaluation points that are most likely to improve the optimization objective function can be determined (or selected).

[0224] The control unit (180) can input a combination of parameters of a selected specific evaluation point into a Bayesian optimization system (100, or a simulator, experimental device, etc.) to calculate or measure the actual objective function value. This process may be a process of obtaining an actual measured value rather than a process of obtaining a predicted value of the model for the selected specific evaluation point. For example, the control unit (180) can input an objective function for the selected specific evaluation point into a Bayesian optimization system (100, or a simulator, experimental device, etc.) and obtain an actual measured value (or actual performance value) from the Bayesian optimization system (100). In the present invention, such an actual measured value may correspond to an evaluation result for the specific evaluation point.

[0225] Additionally, the control unit (180) can expand the initial sample set (or training data set) used in the learning process of the artificial intelligence model (170) by adding a newly acquired specific evaluation point (new parameter combination) and an evaluation result for said specific evaluation point to the initial sample set (or training data set). This can also be understood as a process of updating (or renewing) the initial sample set.

[0226] Furthermore, the updated initial sample set can be used in the update process of the learned artificial intelligence model (170). The control unit (180) can update the learned artificial intelligence model (170) using the updated initial sample set. Alternatively, the control unit (180) can retrain the learned artificial intelligence model (170) using the updated initial sample set. In this case, during retraining, the estimated prior distribution for each parameter is reflected, and the density function of the KDE can also be adaptively updated according to the newly added data. That is, in the present invention, during the retraining process, the newly added evaluation point reduces the uncertainty region of the model and improves the overall prediction accuracy.

[0227] Finally, the control unit (180) can calculate an optimal solution for multiple parameters (or an optimal solution for multiple parameters, an optimal solution for a combination of different multiple parameters, an optimal solution for a combination of different multiple parameters, etc.) based on an update of the artificial intelligence model. Here, the optimal solution may include an optimal combination of parameters calculated from multiple parameters (or from a combination of different multiple parameters). In the present invention, the optimal solution may also be referred to as an “optimal solution” or “optimal combination of parameters,” etc. That is, the control unit (180) can perform Bayesian optimization based on a learned prediction model to proceed with optimization for multiple parameters, and calculate an optimal solution for multiple parameters based on the optimization result (or based on the optimization result).

[0228] Meanwhile, the Bayesian optimization method (or process) according to the present invention described above can be performed iteratively depending on whether a pre-set convergence criterion is satisfied.

[0229] Here, the pre-set convergence criteria may be related to determining whether the search has progressed sufficiently enough to produce an optimal solution. For example, the pre-set convergence criteria may include at least one of i) an improvement limit criterion (e.g., if the improvement range of the objective function value during the last k iterations is below a certain value, the optimal solution candidate is no longer updated and the process is terminated), ii) an acquisition function criterion (e.g., if the maximum value of the acquisition function falls below a certain threshold, it is determined that there is no expected gain to be obtained through further evaluation and the process is terminated), iii) an evaluation budget criterion (e.g., if the maximum allowable number of evaluations (number of experiments / simulations)) is reached, the optimal candidate up to that point is calculated as the optimal solution), and iv) a prediction variance (or uncertainty) criterion of the artificial intelligence model (170) (e.g., if the prediction variance decreases below a threshold in the entire design space, it is determined that sufficient information has been obtained for the entire search space and the process is terminated).

[0230] If the control unit (180) determines that the preset convergence criteria are not satisfied (or are not satisfied), it can repeatedly perform the process of performing KDE for each parameter to estimate individual prior distributions → sampling from the estimated prior distributions → training an artificial intelligence model using the sampled data → selecting the next evaluation point using an acquisition function → performing an evaluation on the selected evaluation point and updating the model → calculating the optimal solution. In the present invention, this process can be performed N times (or N cycles, N iterations, etc.) until the preset convergence criteria are satisfied.

[0231] Meanwhile, in the present invention, the estimated prior distribution for each parameter can be selectively maintained and / or updated (or updated).

[0232] The control unit (180) can selectively update only the prior distribution of the changed parameters when only some of the parameters among a plurality of parameters having different characteristics are changed. Here, selectively updating only the KDE of the changed parameters means a method of updating (or updating) only the probability density estimation (i.e., prior distribution estimation) of the changed parameters without needing to recalculate the distribution of all parameters by using a structure that independently models the prior distribution (or probability distribution) of each parameter in the Bayesian optimization process according to the present invention using an individual kernel density estimation (KDE) technique (or method).

[0233] Specifically, the control unit (180) can re-perform kernel density estimation (KDE) only for specific parameters that have changed values ​​when only some of the parameters (or sets of parameters) having different characteristics are changed during the Bayesian optimization process according to the present invention. That is, the parameters that have not changed can reuse the prior distribution estimation results calculated initially as they are, and the prior distribution can be re-estimated only for specific parameters that have changed by selectively reflecting newly added data (or samples). Here, "change" may mean a case where there is a possibility that the distribution estimation of specific parameter values ​​may change due to new evaluation results (e.g., evaluation results for selected specific evaluation points), experimental data, or external input (input of user terminal (10)).

[0234] In one embodiment, the prior distribution (or information about the prior distribution) estimated for each of a plurality of parameters using a pre-set probability distribution estimation technique may be stored in a storage unit (140, or memory). Each of the plurality of parameters and the prior distribution for each of the plurality of parameters may be matched with each other and stored in the storage unit (140). The control unit (180) may specify (or identify) the changed parameter among the plurality of parameters stored in the storage unit (140) and re-perform kernel density estimation (or probability distribution estimation) only for the specified parameter. The kernel density re-estimation is performed based on existing data (initial sample) + new data (selected specific evaluation point + evaluation result for the specific evaluation point), and the bandwidth (h) and kernel function (K) may be applied according to the characteristics of the corresponding parameter. For the remaining parameters excluding the specified parameter, the existing prior distribution estimation results may be used as is. Through this, the computational cost required to recalculate the prior distribution (KDE) for all parameters from scratch is reduced, and computational efficiency can be maximized by avoiding unnecessary redundant calculations. The process of selectively updating such prior distributions can be carried out during the process of updating the artificial intelligence model (170) based on the evaluation results for a specific evaluation point. Afterwards, the partially updated KDE results and the unchanged KDE results can be integrated and reflected in the sampling process and the model training process.

[0235] In this regard, when the artificial intelligence model (170) is updated based on the evaluation results for a specific evaluation point and the Bayesian optimization process according to the present invention is performed repeatedly, the control unit (180) can perform sampling from each of the prior distribution for the updated specific parameter and the prior distribution for the remaining parameters different from the specific parameter among the plurality of parameters stored in the storage unit (140), and can learn and update the artificial intelligence model (170) using the sampled values.

[0236] In one embodiment, let us assume that after the Bayesian optimization process according to the present invention up to a specific iteration is terminated, the Nth iteration of the Bayesian optimization process according to the present invention is carried out. The control unit (180) can obtain the value of the updated specific parameter and the value of each of the remaining parameters different from the specific parameter by performing sampling from the prior distribution for the updated specific parameter and the prior distribution for the remaining parameters different from the specific parameter, respectively, among the plurality of parameters stored in the storage unit (140). Then, the control unit (180) can generate candidate evaluation points using the value of the updated specific parameter and the value of each of the remaining parameters different from the specific parameter. After the candidate evaluation points are generated, the control unit (180) can select at least one specific evaluation point among the candidate evaluation points to be evaluated using a learned prediction model and a specific function, and perform an evaluation on the selected specific evaluation point. Finally, the control unit (180) can update the learned prediction model based on the evaluation result for the specific evaluation point and calculate the optimal solution for the plurality of parameters based on the update to the learned prediction model.

[0237] In the present invention, performing optimization can be understood as calculating an optimal solution through optimization of optimization parameters. Accordingly, the optimal solution calculated can also be understood as a set of optimal values ​​for all optimization target parameters input by the user (or a final parameter vector, a combination of optimal parameters, an optimized parameter set, etc.). The optimal solution may be configured to include the optimal value of each of a plurality of parameters (or a combination of optimal values).

[0238] As such, the present invention can significantly reduce the amount of computation compared to re-estimating the entire joint distribution by selectively updating (or recalculating) only the prior distribution of the changed parameters. In particular, the iteration speed can be accelerated by reducing unnecessary calculations in high-dimensional problems. Furthermore, in cases of frequent environmental changes or data inflows, real-time optimization is possible by immediately updating only the relevant parameters (e.g., immediately reflecting changes in process conditions to variables to immediately search for improved candidates in the next iteration). Additionally, since the distribution estimation of the changed parameters is maintained in an up-to-date state, the subsequent sampling process better reflects the actual data distribution. Consequently, unnecessary searches are reduced and search efficiency is increased, which can lead to a faster speed in reaching the global optimal value (optimal solution).

[0239] Meanwhile, as described above, the combination optimization system (100) according to the present invention may be linked with an agent server (AS) to provide various services or functions provided in the present invention to the user. The combination optimization system (100) may also include at least one agent (EX: artificial intelligence agent (AI agent)). Furthermore, the agent server may be included and configured in the combination optimization system (100) according to the present invention.

[0240] In the present invention, the agent can interact with the user through a chatbot (chatbot, or conversational artificial intelligence, conversational agent, etc.) linked with the combination optimization system (100). At this time, at least one of the main agent and the sub-agent may be implemented in the form of a chatbot. For example, when the main agent interacts with the user, the main agent may receive user input or provide information (or data) to the user through the chatbot. However, the implementation form of the chatbot in the present invention is merely an example, and it is obvious that it can be implemented in various forms other than a chatbot.

[0241] First, in the present invention, a process of receiving an optimization target parameter to be optimized through a conversation with a user via a user terminal (10) can be carried out in an agent. The agent provides a question related to the optimization target parameter to the user terminal (10) so that the optimization target parameter is specified, and can specify the optimization target parameter based on the answer received from the user terminal (10).

[0242] For example, as illustrated in FIG. 8, the agent can provide guidance information (e.g., “Hello, user :) How can I help you?” 801) to the user terminal (10) through a chatbot (800) that interacts with the combination optimization system (100) in order to receive user input regarding the parameters to be optimized.

[0243] And, the agent can receive optimization target parameters input from the user terminal (10). For example, in one area of ​​the user terminal (10, or service page) where the chatbot (800) is provided, an area for receiving user input (or an input receiving area, 810) may be provided. This input receiving area (810) may include at least one of a text input area configured to receive user input via text, a graphic object (or a first graphic object, 810a) associated with a function for receiving data in an image (or video) format (or format), a graphic object (or a second graphic object, 810b) associated with a function for receiving data in a file and / or document format (e.g., Excel, Word, XML, JSON, YAML, Gerber, ODB, CAD file, etc.), and a graphic object (or a third graphic object, 810c) associated with a function for receiving user input.

[0244] The agent may receive a document (802a) and text data (e.g., “The attached file contains parameters set inside the appearance inspector, please find the optimal combination,” 802) corresponding to user input from the user terminal (10) based on the selection of a third graphic object (810c) included in the input receiving area (810). This document (802a) may include at least one of the optimization target parameter requested by the user through the text data (802) and information regarding said optimization target parameter (e.g., the search range of the optimization target parameter). For example, the optimization target parameter may include multiple parameters having different characteristics. In this case, the document (802a) may include at least one of the multiple parameters and the search range of each of said multiple parameters. At this time, the text data (802) may also be understood as a user query corresponding to the user request, and receiving the document (802a) and text data (802) may also be understood as receiving an “input value related to the optimization target parameter.”

[0245] Subsequently, the agent can transmit the document (802a) and text data (802) received through the chatbot (800) to the agent server (AS). The agent server (AS) performs an analysis on the document (802a) and text data (802), and based on the analysis results, can extract at least one of a plurality of parameters to be optimized, a search range for each of the plurality of parameters, and a value for each of the plurality of parameters (parameter value).

[0246] At this time, the agent server (AS) can specify target information based on the document (802a) and text data (802). Then, the agent server (AS) can generate a work plan to optimize the target parameters based on the specified target information. According to the work plan, the agent can perform optimization based on the parameter optimization target information by cooperating with at least one of the probability distribution estimation unit (150), data processing unit (160), artificial intelligence model (170), and simulator of the combination optimization system (100). Such an agent may include a main agent that generates the work plan and a plurality of sub-agents that cooperate with the main agent and are respectively linked to any one of the probability distribution estimation unit (150), data processing unit (160), artificial intelligence model (170), and simulator.

[0247] And, the main agent transmits a work command to at least one of the multiple sub-agents according to the work plan, and the at least one sub-agent that receives the work command can perform parameter optimization according to the parameter optimization goal information according to the work command.

[0248] More specifically, the agent server (AS) may select (or specify) each of the agents among a plurality of agents (A1, A2, A3) to “perform optimization on the parameters to be optimized and calculate (or extract, derive, generate, etc.) the optimal solution for the parameters to be optimized.” For example, the agent server (AS) may select the first agent (A1), the second agent (A2), and the third agent (A3) among a plurality of agents (or sub-agents (A1, A2, A3, etc.)) specialized in different functions to optimize the parameters to be optimized included in the document (802a) based on the document (802a) and text data (802) entered by the user.

[0249] In this case, each of the selected multiple agents (A1, A2, A3) can understand the user intent contained in the document (802a) and text data (802) and perform customized optimization according to the user intent.

[0250] In the present invention, each of the plurality of agents (A1, A2, A3) selected according to the work plan can perform a search for a plurality of candidate solutions generated according to input values ​​using a first optimization algorithm based on target information, generate result data based on the results of the search for the plurality of candidate solutions, train an artificial intelligence model (170) using the generated result data, and perform optimization for the target parameters based on the trained artificial intelligence model (170) and a pre-set second optimization algorithm.

[0251] At this time, the artificial intelligence model (170) may be a model that has been trained in advance. For example, the artificial intelligence model (170) may be a model trained based on result data generated by performing a search on parameters input by users in advance. Alternatively, the artificial intelligence model (170) may be a model that is set (or adjusted) based on result data generated according to user input (e.g., parameters to be optimized) so as to perform optimization on parameters to be optimized input by the user. That is, the artificial intelligence model (170) may be set to perform optimization in accordance with user input input in real time.

[0252] In one embodiment, the first agent (A1) can generate a plurality of candidate solutions based on input values ​​related to optimization target parameters entered by a user using a preset first optimization algorithm, perform an evaluation on each of the plurality of candidate solutions to produce an evaluation result for each of the plurality of candidate solutions, and select an optimal candidate solution among the plurality of candidate solutions that satisfies a preset criterion based on the evaluation result for each of the plurality of candidate solutions. Furthermore, the first agent (A1) can update at least some of the plurality of candidate solutions using a preset computation method, perform an evaluation on the updated candidate solutions to produce an evaluation result for the updated candidate solutions, and select an optimal candidate solution among the updated candidate solutions that satisfies a preset criterion based on the evaluation result for the updated candidate solutions. Such a first agent (A1) can perform a search using the first optimization algorithm N times, corresponding to a preset number of iterations N, and generate result data based on the search results performed N times. At this time, the first agent (A1) can store the generated result data in a storage unit (140) or transmit (or deliver) it to the second agent (A2).

[0253] In another embodiment, the second agent (A2) can estimate a prior distribution for a parameter to be optimized using a pre-set probability distribution estimation technique, and can train (or set) an artificial intelligence model (170) using at least one of the prior distribution and result data for the parameter to be optimized. The second agent (A2) can estimate a prior distribution for each of a plurality of parameters having different characteristics corresponding to the parameter to be optimized, and can train an artificial intelligence model (170) using at least one of the prior distribution and result data for each of the plurality of parameters.

[0254] In another embodiment, the third agent (A3) can perform optimization on the target parameter based on the learned artificial intelligence model (170) and the pre-configured second optimization algorithm. The third agent (A3) can calculate an optimal solution for the target parameter by performing optimization on the search space defined based on the target parameter input by the user, based on the learned artificial intelligence model. In this case, the third agent (A3) can update the learned artificial intelligence model based on the parameter combination specified from the search space, and perform optimization on the target parameter based on the update to the learned artificial intelligence model.

[0255] In the present invention, performing optimization can be understood as calculating an optimal solution through optimization of optimization parameters. Accordingly, the optimal solution calculated can also be understood as a set of optimal values ​​for all optimization target parameters input by the user (or a final parameter vector, a combination of optimal parameters, an optimized parameter set, etc.). The optimal solution may be configured to include the optimal value of each of a plurality of parameters (or a combination of optimal values).

[0256] Finally, the agent server (AS) can provide the optimal solution for the optimization target parameters entered by the user to the user terminal (10). For example, as illustrated in FIG. 9, the agent server (AS) can provide the user terminal (10) with a response (or answer, 820) that includes at least one of the optimal solution (821) calculated based on the search and optimization results of each of the plurality of agents (A1, A2, A3) and the evaluation result for the optimal solution (821). Here, the optimal solution (821) can also be understood as a combination of optimal parameters.

[0257] Additionally, the response (820) may further include a graphic object (820a) associated with a provision function for other solutions discovered during the process of calculating the optimal solution (821). For example, as illustrated in FIGS. 9 and 10, the agent server (AS) may provide the user terminal (10) with a response (830) that includes at least one of other solutions (831, 832) discovered during the process of calculating the optimal solution (821) and an evaluation result for each of the other solutions (831, 832), based on the selection of the graphic object (820a) from the user terminal (10).

[0258] In this way, the present invention can understand the intent (or context) of user input to select an appropriate agent, and through the selected agent, generate and provide customized results according to the user's intent to the user.

[0259] As described above, according to the combinatorial optimization method and system of the present invention, by combining the global search capability of an evolutionary algorithm (EA) and the precise local search capability of Bayesian optimization (BO), it is possible to extensively explore the initial solution space and perform fine parameter adjustments. Accordingly, the present invention effectively prevents the cold start problem, in which the prediction or search performance of a model (or algorithm) deteriorates due to a lack of initial data, and can derive the optimal solution to the problem quickly and accurately. In other words, the present invention can achieve stable and high-performance optimization even in situations where initial data is insufficient.

[0260] Furthermore, according to the combinatorial optimization method and system of the present invention, various initial solutions are efficiently generated through the search of an evolutionary algorithm and utilized in Bayesian optimization, thereby enabling efficient and accurate parameter optimization even with limited data. Through this, the present invention can maintain high optimization performance while minimizing the consumption of computational (or operation) resources, and can effectively search a complex multidimensional parameter space.

[0261] Furthermore, according to the combinatorial optimization method and system of the present invention, for a plurality of interacting parameters (or variables), optimization reflecting the interaction effects between multiple parameters is possible by learning and adjusting candidate combinations generated using an evolutionary algorithm through Bayesian optimization. Through this, the present invention can simultaneously improve the speed and accuracy of optimization even in environments requiring real-time data, such as parameter tuning of machine learning models, industrial process optimization, and visual inspection machines. Accordingly, the present invention can provide efficient and accurate optimization results in various fields such as parameter tuning of machine learning models, industrial process optimization, and visual inspection machines. Consequently, the present invention converges to an optimal solution faster and more stably than when using a single algorithm alone, thereby providing excellent performance and resource efficiency across complex combinatorial optimization problems and enabling the advancement of parameter combination optimization.

[0262] Meanwhile, according to the combinatorial optimization method and system of the present invention, optimization performance can be improved by estimating the individual distribution for each of the multiple parameters in an environment where multiple parameters having different characteristics are mixed. That is, by processing each parameter individually, the present invention can prevent performance degradation in high-dimensional problems and mitigate the curse of dimensionality.

[0263] Furthermore, according to the combinatorial optimization method and system of the present invention, an artificial intelligence model can be trained based on at least one of the result data generated using a prior distribution and an evolution algorithm for each of a plurality of parameters, and parameter optimization can be performed based on the trained artificial intelligence model. Through this, the present invention can minimize the total search space and concentrate search resources on a valid and high-performance space, thereby reducing unnecessary evaluations and improving the optimization convergence speed.

[0264] Meanwhile, regarding examples to which the present invention can be applied, the combination optimization system (100) of the present invention can be applied to parameter optimization in a semiconductor etching process. For example, a first agent (A1) uses an evolutionary algorithm (EA) to search for multiple process control parameters, such as the temperature, gas flow rate, RF power, and pressure of an etching chamber, and generates result data based on the search results. A second agent (A2) can use the result data to estimate a prior distribution that reflects the interaction effect between each parameter and train a Bayesian Neural Network to construct a probabilistic prediction model for process results (e.g., etching uniformity, critical dimension error, etc.). Subsequently, a third agent (A3) can use the trained model to derive candidate parameter combinations that maximize the Expected Improvement (EI) within the search space, thereby minimizing process deviation and stably maintaining the target quality. Accordingly, in this embodiment, the effect of maximizing optimization efficiency while reducing the number of iterative experiments without process simulation can be achieved.

[0265] In another embodiment, the combination optimization system (100) of the present invention can be applied to tuning inspection algorithm parameters of an industrial vision inspection system. For example, a first agent (A1) generates multiple candidate image data by combining parameters such as lighting brightness, lighting angle, exposure time, and threshold, and evaluates defect detection precision, recall, F1-score, etc. for each candidate solution. Subsequently, a second agent (A2) trains an artificial intelligence model (170) based on the evaluation results to estimate the predicted value and uncertainty of image quality according to parameter changes. A third agent (A3) calculates an optimal parameter set that obtains maximum performance with the minimum number of test samples by performing an Upper Confidence Bound (UCB)-based search that improves search efficiency using the trained model. As a result, in this embodiment, technical effects can be obtained that reduce computational resources compared to conventional manual-based parameter adjustment and shorten the time required to deploy the AI ​​inspection system into the production line.

[0266] In another embodiment, the combination optimization system (100) of the present invention can be applied to optimize real-time motion parameters (e.g., joint torque, velocity, trajectory correction value) of a manufacturing robot. A first agent (A1) receives real-time sensor data (torque, acceleration, temperature, etc.) as input, performs an evolutionary algorithm-based search, and evaluates the energy consumption, error rate, vibration value, etc. of each candidate motion combination. A second agent (A2) combines the evaluation results with the robot's dynamic state model to train a Bayesian optimization model, and a third agent (A3) can perform real-time parameter correction within a control loop using the trained model. Through this, self-adaptive control that maximizes energy efficiency while maintaining the motion stability of the robot becomes possible.

[0267] In another embodiment, the combination optimization system (100) of the present invention can be applied to hyperparameter tuning of a large-scale language model (LLM) or recommendation system. For example, a first agent (A1) searches for hyperparameters such as a learning rate, batch size, embedding dimension, and regularization coefficient using an evolutionary algorithm. A second agent (A2) trains a Gaussian Process Model (GPR) based on the search result data to estimate the performance prediction value of each hyperparameter combination. A third agent (A3) performs optimization by selecting a combination that maximizes the Expected Improvement (EI) function, and by repeating this process, improves the learning efficiency and convergence speed of the model. Through this embodiment, efficient search becomes possible even in a complex deep learning parameter space, and technical effects such as reducing large-scale computation costs can be obtained.

[0268] In another embodiment, the combination optimization system (100) of the present invention can be applied to optimize complex parameters such as sensor weights, time synchronization delays, and noise filter coefficients in a multimodal sensor fusion system. A first agent (A1) generates various sensor combination data through initial exploration, and a second agent (A2) models the correlation between sensors by training a Bayesian neural network based on the error variance of each combination. A third agent (A3) can simultaneously improve sensor fusion accuracy and system response speed by deriving a parameter set in which the prediction variance of the trained model is minimized.

[0269] In addition, the first to third agents (A1 to A3) in this specification are distinguished for convenience of explanation and are named differently according to their functional roles; however, not all agents must be mutually independent or are necessarily limited to three. That is, the concept of the present invention can be equally applied to structures including two or fewer or four or more agents, and the configuration, order, or method of interoperability of each agent may be varied depending on the implementation environment.

[0270] Furthermore, according to the combination optimization method and system of the present invention, computational costs are reduced by estimating the prior distribution for each of a plurality of parameters having different characteristics, and individual learning rates or bandwidth adjustments are possible to suit the characteristics of each parameter.

[0271] Furthermore, according to the combination optimization method and system of the present invention, each parameter can be processed simultaneously according to the characteristics of each of the multiple parameters, and accurate estimation is possible even within a wide optimization range of each parameter.

[0272] Furthermore, according to the combinatorial optimization method and system of the present invention, when only some parameters among a plurality of parameters having different characteristics are changed, only the prior distribution of the changed parameters can be selectively updated. Through this, the present invention can improve the efficiency of incremental learning and can apply different bandwidths to each parameter, thereby increasing the accuracy of distribution estimation according to the characteristics of each variable. In addition, distribution information of some parameters can be reused from similar optimization problems to be quickly applied to new problems, and the distribution of related parameters can be immediately updated whenever new data is received, thereby enabling real-time response to environmental changes. Accordingly, since the present invention only requires storing the distribution information of each parameter, memory can be used efficiently to save storage space.

[0273] As such, the present invention efficiently processes complex mixed variables and enables accurate estimation even within a wide optimization range for each variable. This method can provide improved performance in high-dimensional problems, effective processing of mixed variables, and increased computational efficiency. In particular, the present invention can be usefully applied to real industrial environments where interactions between variables are complex and the range of influence of each variable is wide, such as factory automation or complex system optimization.

[0274] Meanwhile, the present invention described above can be implemented based on a quantum computer. The present invention implemented based on a quantum computer may include a qubit-based quantum processor and quantum memory, and may include software and hardware interfaces optimized for quantum computation.

[0275] Quantum processors in quantum computers utilize qubits to efficiently process complex operations through parallel computation, quantum entanglement, and quantum superposition, which cannot be performed by the binary bits of classical computers. Quantum processors process data using quantum gates and can provide exponential speed improvements for specific problems.

[0276] Meanwhile, the present invention described above can be implemented as a program that is executed by one or more processes on a computer and can be stored on a computer-readable medium (or recording medium).

[0277] Furthermore, the present invention described above can be implemented as computer-readable code or instructions on a medium on which a program is recorded. That is, the present invention can be provided in the form of a program.

[0278] Meanwhile, computer-readable media include all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

[0279] Furthermore, the computer-readable medium may be a server or cloud storage that includes a storage and is accessible to an electronic device via communication. In this case, the computer may download the program according to the present invention from the server or cloud storage via wired or wireless communication.

[0280] A computer program may reach the system (100) through various suitable transmission mechanisms. The transmission mechanism may be, for example, a computer-readable storage medium, a computer program product, a memory device, a recording medium such as a CD-ROM or DVD, or a product that tangibly embodies the computer program. The transmission mechanism may be a signal configured to reliably transmit the computer program through air or an electrical connection. The system (100) may propagate or transmit the computer program as a computer data signal.

[0281] Furthermore, references to 'computer-readable storage media,' 'computer program products,' 'computer programs embodied in a tangible form,' etc., or to 'controller,' 'computer,' 'processor,' etc., should be understood to include not only computers with various architectures such as single / multi-processor architectures and sequential (Von Neumann) / parallel architectures, but also specialized circuits such as Field-Programmable Gate Arrays (FPGAs), Application Specific Circuits (ASICs), signal processing units, and other devices. References to computer programs, instructions, code, etc., should be understood to include software for programmable processors or firmware, such as programmable content for hardware devices, whether it is instructions for a processor or configuration settings for a fixed-function device, gate array, or programmable logic device.

[0282] Furthermore, in the present invention, the computer described above is an electronic device equipped with a processor, namely a CPU (Central Processing Unit), and no special limitations are placed on its type.

[0283] Meanwhile, the above detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention should be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention.

Claims

1. In a computerized method comprising the following, A step of specifying input values ​​related to parameters to be optimized; A step of performing a search for a plurality of candidate solutions generated according to the input value using a pre-configured first optimization algorithm; A step of generating result data based on the search results for the plurality of candidate solutions; A step of training an artificial intelligence model using the above-mentioned generated result data; and A combination optimization method characterized by performing optimization on the optimization target parameter based on the above-mentioned learned artificial intelligence model and a pre-set second optimization algorithm.

2. In Paragraph 1, The above input value is, It includes at least one of the number of candidate solutions including the optimization target parameter, the number of optimization target parameters included in the candidate solutions, and the search range of the optimization target parameter. The step of performing the above search is, A step of repeatedly performing the search using the first optimization algorithm based on a preset number of iterations for the first optimization algorithm, and In the step of generating the above result data, A combination optimization method characterized by generating result data based on the search results repeatedly performed according to the number of repetitions set above.

3. In Paragraph 2, In the step of performing the above search, The search using the first optimization algorithm is performed N times, corresponding to the number of repetitions N set above, and In the step of generating the above result data, A combination optimization method characterized by generating result data based on the search results performed N times.

4. In Paragraph 2, In the step of performing the above search, Using the above first optimization algorithm, the plurality of candidate solutions are generated according to the above input value, and In order to select an optimal candidate solution satisfying a preset criterion from the plurality of candidate solutions, a search is performed for each of the plurality of candidate solutions, and In the step of generating the above result data, A combination optimization method characterized by generating result data based on the search results for each of the plurality of candidate solutions.

5. In Paragraph 4, In the step of generating the above plurality of candidate solutions, A combination optimization method characterized by generating a plurality of candidate solutions based on the search range of the optimization target parameters included in the input values.

6. In Paragraph 4, In the step of performing the above search, An evaluation is performed for each of the above-mentioned plurality of candidate solutions, and an evaluation result for each of the above-mentioned plurality of candidate solutions is produced. Based on the evaluation results for each of the plurality of candidate solutions, the optimal candidate solution among the plurality of candidate solutions that satisfies the preset criteria is selected, and The above result data is, A combination optimization method characterized by including at least one of the plurality of candidate solutions and an evaluation result for each of the plurality of candidate solutions.

7. In Paragraph 6, In the step of performing the above search, Using a preset computational technique, at least some of the above multiple candidate solutions are updated, and A combination optimization method characterized by performing a search on the updated candidate solutions to select an optimal candidate solution that satisfies the previously set criteria from the updated candidate solutions.

8. In Paragraph 7, An evaluation of the above-mentioned updated candidate solution is performed to produce an evaluation result for the above-mentioned updated candidate solution, and Based on the evaluation results of the above-mentioned updated candidate solutions, the above-mentioned optimal candidate solution satisfying the above-mentioned preset criteria is selected among the above-mentioned updated candidate solutions, and The above result data is, A combination optimization method characterized by further including at least one of the above-mentioned updated candidate solution and the evaluation result for the above-mentioned updated candidate solution.

9. In Paragraph 1, The step of performing the above search is, A step of searching for at least one optimal candidate solution for each iteration based on a preset number of iterations for the first optimization algorithm, and A combinatorial optimization method characterized in that the above optimal candidate solution is included in the above result data and used for training the above artificial intelligence model.

10. In Paragraph 9, In the above result data, A combination optimization method characterized by including at least one of the optimal candidate solution, an evaluation result for the optimal candidate solution, and a search range for the optimal candidate solution.

11. In Paragraph 1, To train the artificial intelligence model, the method further includes the step of storing the result data in a specified repository. In the step of performing the above optimization, A combination optimization method characterized by performing optimization on a search space defined based on the above-mentioned optimization target parameters.

12. In Paragraph 11, In the step of training the above artificial intelligence model, To perform optimization on the above-mentioned optimization target parameters, the artificial intelligence model is trained using the result data stored in the above-mentioned specified repository, and In the step of performing the above optimization, A combinatorial optimization method characterized by performing optimization on the search space based on the above-mentioned learned artificial intelligence model.

13. In Paragraph 11, In the step of performing the above optimization, A combination optimization method characterized by performing optimization on the target parameter through at least one parameter combination specified from the search space using the above-mentioned learned artificial intelligence model and a specific function.

14. In Paragraph 13, The above-mentioned trained artificial intelligence model is, A combination optimization method characterized by predicting at least one of a predicted value and a predicted variance for at least one combination of parameters.

15. In Paragraph 13, In the step of performing the above optimization, Update the learned artificial intelligence model based on the parameter combination specified from the above search space, and A combination optimization method characterized by performing optimization on the optimization target parameter based on the update to the learned artificial intelligence model.

16. In Paragraph 1, In the step of performing the above optimization, A combinatorial optimization method characterized by performing optimization on the target parameter based on the updated artificial intelligence model and the second optimization algorithm to produce an optimal solution for the target parameter.

17. In Paragraph 1, The method further includes the step of estimating a prior distribution for the optimization target parameter using a pre-established probability distribution estimation technique, and In the step of training the above artificial intelligence model, A combination optimization method characterized by training the artificial intelligence model using at least one of the prior distribution for the optimization target parameter and the result data.

18. In Paragraph 17, The above optimization target parameters include a plurality of parameters having different characteristics, and In the step of estimating the prior distribution mentioned above, Using the above-mentioned probability distribution estimation technique, the prior distribution for each of the above-mentioned plurality of parameters is estimated, and In the step of training the above artificial intelligence model, A combination optimization method characterized by training the artificial intelligence model using at least one of the prior distribution for each of the plurality of parameters and the result data.

19. A system comprising memory configured to store executable instructions and one or more processors configured to perform operations by executing one or more instructions, The above system is, Specify input values ​​related to the parameters to be optimized, and Using a pre-configured first optimization algorithm, a search for a plurality of candidate solutions generated according to the input value is performed, and Result data is generated based on the search results for the above multiple candidate solutions, and Using the result data generated above, an artificial intelligence model is trained, and A combination optimization system characterized by performing optimization on the optimization target parameter based on the above-mentioned learned artificial intelligence model and a pre-set second optimization algorithm.

20. A program that is executed by one or more processes in an electronic device and stored on a computer-readable recording medium, The above program is, A step of specifying input values ​​related to parameters to be optimized; A step of performing a search for a plurality of candidate solutions generated according to the input value using a pre-configured first optimization algorithm; A step of generating result data based on the search results for the plurality of candidate solutions; A step of training an artificial intelligence model using the above-mentioned generated result data; and A program stored on a computer-readable recording medium characterized by including instructions for performing a step of optimizing the optimization target parameter based on the above-mentioned learned artificial intelligence model and a pre-set second optimization algorithm.

Images