A method for improving accuracy of artificial intelligence and electronic apparatus thereof

Abstract

According to the present disclosure, a method for improving the accuracy of artificial intelligence (AI) of an electronic device comprises: designing an AI based on a Monte-Carlo simulation algorithm; setting a region of interest for executing a local Monte-Carlo simulation within a vulnerable region where the AI ​​is determined to be vulnerable among the application regions of the AI; executing a local Monte-Carlo simulation for the region of interest based on the AI; and confirming information regarding the improved accuracy of the AI ​​based on the execution of the local Monte-Carlo simulation for the region of interest, wherein the improved accuracy is determined based on the fact that the deep learning data obtained through the execution of the local Monte-Carlo simulation for the region of interest is additionally reflected in the deep learning data obtained through the execution of the Monte-Carlo simulation for the application region.

Description

Technology Field

[0001] The present invention relates to a method and apparatus for improving the accuracy of artificial intelligence, and more specifically, to a method and electronic apparatus for improving the accuracy of artificial intelligence through the execution of local simulations of vulnerable areas of artificial intelligence. Background Technology

[0003] In the design of AI (Artificial Intelligence) systems, when it is impossible to simulate numerous cases for each stage, the Monte Carlo algorithm—a random sampling method—is frequently used to approach approximations. As such, the Monte Carlo algorithm is fundamentally widely utilized in AI development to find the optimal solution within big data.

[0004] Taking Go AI services as an example, Monte-Carlo Tree Search (MCTS) assumes that both the player and the opponent have the same policy network in the current state. It runs multiple simulations to find the best next move and selects the move identified with the highest frequency. This method is known to produce near-optimal results if sufficient simulations can be run using high-performance equipment, even without exploring the entire tree. In practice, Google supports parallel processing and allocates a large number of CPUs and GPUs for computation to ensure sufficient simulations.

[0005] However, the MCTS algorithm is inefficient because it relies on random sampling and requires a vast amount of trial-and-error. Consequently, current AI services based on the MCTS algorithm are difficult to implement without supporting hardware specifications, such as high-performance GPUs. This implies that there is a disadvantage in that high costs and significant time are required to fully implement AI services.

[0006] For example, in the case of Go AI, in the fourth game—the only victory for Lee Sedol among the five matches between him and AlphaGo—Lee Sedol initiated a "fight for survival" as early as the 78th move, when not many moves had been made yet. This involved a physical struggle in the center, where the number of possible liberties was at its maximum of 4, with stones placed almost completely densely and without gaps. As the number of possible liberties in the center of the fight for survival became extremely large, consisting mainly of multiples of 4 such as "4×4×4×...", this significantly weakened the strength of AlphaGo's MCTS-based AI, which determines the breadth of its move selection and the depth of its next move progression, thereby enabling Lee Sedol to win. To use a simpler analogy, the central area of ​​the board is easily visible to the human eye of Lee Sedol, allowing him to easily predict the possibilities in the fight for survival; however, to AlphaGo's vision based on MCTS, the center of the board appears as a deep region, like a black hole, that is difficult to see. Ultimately, in the case of Go AI, the central area is a region that is difficult to see due to the large number of possible moves in the early stages; therefore, it can be said that it is vulnerable to tactics that induce a large number of early-game fighting strategies in this area.

[0007] As a method to address the fundamental problem of AI dealing with a 'very large number of possibilities' as described above, neural network structures such as value networks and policy networks have traditionally been introduced and applied to AlphaGo and other Go AIs. However, even with the introduction of these neural network concepts, Go AIs still suffer from low accuracy, such as failing to properly predict win probabilities for large-scale fighting in vulnerable areas like the center. The problem to be solved

[0009] According to the method of the present invention, an electronic device can improve the accuracy of artificial intelligence by performing a local simulation of a vulnerable area of ​​artificial intelligence.

[0010] The technical problems to be solved by the present invention are not limited to the technical problems mentioned above, and the mentioned

[0011] Other technical problems not mentioned below will be clearly understood by those skilled in the art to which the present invention belongs from the description below. means of solving the problem

[0013] Various embodiments can provide a method for improving artificial intelligence (AI) accuracy and an electronic device.

[0014] A method for improving the accuracy of artificial intelligence (AI) of an electronic device according to various embodiments comprises: a step of designing an AI based on a Monte-Carlo simulation algorithm; a step of setting a region of interest for executing a local Monte-Carlo simulation within a vulnerable region where the AI ​​is determined to be vulnerable among the application regions of the AI; a step of executing a local Monte-Carlo simulation for the region of interest based on the AI; and a step of confirming information regarding the improved accuracy of the AI ​​based on the execution of the local Monte-Carlo simulation for the region of interest, wherein the improved accuracy can be determined based on the fact that the deep learning data obtained through the execution of the local Monte-Carlo simulation for the region of interest is additionally reflected in the deep learning data obtained through the execution of the Monte-Carlo simulation for the application region.

[0015] In an exemplary embodiment, the AI ​​may be configured to apply a Latin Hypercube algorithm instead of the Monte Carlo simulation algorithm based on conditions related to uniformity and efficiency.

[0016] In an exemplary embodiment, when the AI ​​corresponds to a Go AI, an M*M line region of interest is set within a central part of the entire N*N line Go board area where the AI ​​is determined to be weak, and the enhanced accuracy is determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation on the M*M line region of interest is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation on the entire N*N line Go board area, and M may be smaller than N.

[0017] In an exemplary embodiment, when the AI ​​corresponds to a medical diagnosis AI, a medical diagnosis interest region is set in which the medical diagnosis region is judged to have a high rate of medical error judgment by the AI, and the enhanced accuracy can be determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation on the medical diagnosis interest region is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation on the medical diagnosis region.

[0018] In an exemplary embodiment, if it is determined that the AI ​​corresponds to an application error state with respect to a specific area among the application areas, the AI ​​may be configured to exclude the application result of the AI ​​for the specific area based on a processing algorithm pre-set in the AI, and to obtain only the application result of the AI ​​for the remaining area excluding the specific area among the application areas.

[0019] In an exemplary embodiment, the method for improving AI accuracy may further include: a step of setting a plurality of regions of interest having different sizes within the vulnerable region; a step of determining a respective improved accuracy for the AI ​​based on the execution of a local Monte Carlo simulation for each region of interest included in the plurality of regions of interest; and a step of setting information including the size of each region of interest and each improved accuracy.

[0020] In an exemplary embodiment, the AI ​​accuracy enhancement method may further include: a step of determining the degree of computational load determined when executing a local Monte Carlo simulation for each of the plurality of regions of interest; and a step of determining a region of interest of an optimal size for the performance of the AI ​​among the plurality of regions of interest based on each of the enhanced accuracy and each degree of computational load.

[0021] An electronic device for performing a method to improve the accuracy of artificial intelligence (AI) according to various embodiments comprises: a processor; and one or more memories for storing one or more instructions, wherein, when the one or more instructions are executed, the processor controls the processor to perform: a step of designing an AI based on a Monte-Carlo simulation algorithm; a step of setting a region of interest for executing a local Monte-Carlo simulation within a vulnerable region where the AI ​​is determined to be vulnerable among the application regions of the AI; a step of executing a local Monte-Carlo simulation for the region of interest based on the AI; and a step of confirming information regarding the improved accuracy of the AI ​​based on the execution of the local Monte-Carlo simulation for the region of interest, wherein the improved accuracy may be determined based on the fact that the deep learning data obtained through the execution of the local Monte-Carlo simulation for the region of interest is additionally reflected in the deep learning data obtained through the execution of the Monte-Carlo simulation for the application region.

[0022] In an exemplary embodiment, the AI ​​may be configured to apply a Latin Hypercube algorithm instead of the Monte Carlo simulation algorithm based on conditions related to uniformity and efficiency.

[0023] In an exemplary embodiment, when the AI ​​corresponds to a Go AI, an M*M line region of interest is set within a central part of the entire N*N line Go board area where the AI ​​is determined to be weak, and the enhanced accuracy is determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation on the M*M line region of interest is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation on the entire N*N line Go board area, and M may be smaller than N.

[0024] In an exemplary embodiment, when the AI ​​corresponds to a medical diagnosis AI, a medical diagnosis interest region is set in which the medical diagnosis region is judged to have a high rate of medical error judgment by the AI, and the enhanced accuracy can be determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation on the medical diagnosis interest region is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation on the medical diagnosis region.

[0025] In an exemplary embodiment, if it is determined that the AI ​​corresponds to an application error state with respect to a specific area among the application areas, the AI ​​may be configured to exclude the application result of the AI ​​for the specific area based on a processing algorithm pre-set in the AI, and to obtain only the application result of the AI ​​for the remaining area excluding the specific area among the application areas.

[0026] In an exemplary embodiment, the method for improving AI accuracy may further include: a step of setting a plurality of regions of interest having different sizes within the vulnerable region; a step of determining a respective improved accuracy for the AI ​​based on the execution of a local Monte Carlo simulation for each region of interest included in the plurality of regions of interest; and a step of setting information including the size of each region of interest and each improved accuracy.

[0027] In an exemplary embodiment, the AI ​​accuracy enhancement method may further include: a step of determining the degree of computational load determined when executing a local Monte Carlo simulation for each of the plurality of regions of interest; and a step of determining a region of interest of an optimal size for the performance of the AI ​​among the plurality of regions of interest based on each of the enhanced accuracy and each degree of computational load.

[0028] The various embodiments of the present disclosure described above are merely some of the preferred embodiments of the present disclosure, and various embodiments reflecting the technical features of the various embodiments of the present disclosure can be derived and understood by those skilled in the art based on the detailed description to be described below. Effects of the invention

[0030] The present invention has a technical effect in that it can improve the accuracy of artificial intelligence by allowing an electronic device to set a region of interest within a vulnerable area of ​​artificial intelligence and to reflect deep learning data obtained through the execution of a local simulation of the region of interest in the operation of artificial intelligence.

[0031] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0033] FIG. 1 is a diagram illustrating the configuration of an electronic device for performing the method for improving artificial intelligence accuracy of the present disclosure. Specific details for implementing the invention

[0034] Hereinafter, various embodiments of the present disclosure are described in detail with reference to the accompanying drawings so that those skilled in the art can easily practice them. The detailed description disclosed below is intended to explain exemplary embodiments of various examples and is not intended to limit the scope to specific embodiments.

[0035] The terms used in the embodiments have been selected to be as widely used as possible, taking into account their functions in the present disclosure; however, these may vary depending on the intent of those skilled in the art, case law, the emergence of new technologies, etc. Additionally, in specific cases, terms have been arbitrarily selected by the applicant, and in such cases, their meanings will be described in detail in the relevant explanatory section. Therefore, terms used in the present disclosure should be defined not merely by their names, but based on their meanings and the overall content of the present disclosure.

[0036] The following embodiments are combinations of the components and features of various embodiments in a predetermined form. Each component or feature may be considered optional unless otherwise explicitly stated. Each component or feature may be implemented in a form not combined with other components or features. Additionally, various embodiments may be constructed by combining some components and features. The order of operations described in various embodiments may be changed. Some components or features of one embodiment may be included in another embodiment, or may be replaced with corresponding components or features of another embodiment.

[0037] In the description of the drawings, procedures or steps that could obscure the essence of the various embodiments were not described, nor were procedures or steps that could be understood by a person with ordinary knowledge in the relevant technical field described.

[0038] Throughout the specification, when a part is described as "comprising" or "including" a component, it means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. In this document, the singular form of a noun corresponding to an item may be used to include both singular and plural forms, unless otherwise indicated in the specification or clearly contradicted by the context. In this document, phrases such as "A or B," "at least one of A and B," "at least one of A or B," "A, B or C," "at least one of A, B and C," and "at least one of A, B, or C" may each include any one of the items listed together in the corresponding phrase, or any combination thereof. Terms such as "first," "second," or "first" or "second" may be used simply to distinguish a component from other corresponding components and do not limit the components in any other aspect (e.g., importance or order).

[0039] Each component (e.g., module or program) of the components described in this document may include a singular or multiple entities. According to various embodiments, one or more of the components or operations may be omitted, or one or more other components or operations may be added. Additionally, multiple components (e.g., module or program) may be integrated into a single component. In this case, the integrated component may perform one or more functions of each of the components of the multiple components in the same or similar manner as they were performed by the corresponding component among the multiple components prior to integration. According to various embodiments, operations performed by a module, program, or other component may be executed sequentially, in parallel, or repeatedly; one or more of the operations may be executed in a different order; may be omitted; or one or more other operations may be added.

[0040] As used in this document, the terms "module" or "...part" refer to a unit that processes at least one function or operation, and may include a unit implemented by hardware, software, firmware, or a combination thereof.

[0041] Various embodiments of this document may be implemented as software (e.g., a program or application) comprising one or more instructions stored in a storage medium (e.g., memory) readable by a machine. For example, the processor of the machine may call at least one of the one or more instructions stored from the storage medium and execute it. This enables the machine to operate to perform at least one function according to at least one called instruction. One or more instructions may include code generated by a compiler or code that can be executed by an interpreter. The storage medium readable by a machine may be provided in the form of a non-transitory storage medium. Here, 'non-transitory' simply means that the storage medium is a tangible device and does not contain a signal, and does not distinguish between cases where data is stored semi-permanently and cases where it is stored temporarily in the storage medium.

[0042] In addition, specific terms used in various embodiments are provided to aid in understanding the various embodiments, and the use of such specific terms may be modified in other forms within the scope of not departing from the technical concept of the various embodiments.

[0043] FIG. 1 is a diagram illustrating the configuration of an electronic device for performing the method for improving artificial intelligence accuracy of the present disclosure.

[0044] Referring to FIG. 1, the electronic device (100) may include a processor (110) and a memory (120) according to one embodiment. The electronic device (100) illustrated in FIG. 1 is illustrated only with components related to the present embodiment, and it will be understood by those skilled in the art related to the present embodiment that other general-purpose components may be included in addition to the components illustrated in FIG. 1.

[0045] For example, an electronic device (100) may include a communication device comprising one or more transceivers, an input unit, and an output unit. The communication device is a device for performing wired / wireless communication and can communicate with an external electronic device. The external electronic device may be a terminal or a server. In addition, communication technologies used by the communication device may include GSM (Global System for Mobile communication), CDMA (Code Division Multi Access), LTE (Long Term Evolution), 5G, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Bluetooth, RFID (Radio Frequency Identification), Infrared Data Association (IrDA), ZigBee, NFC (Near Field Communication), etc. The input unit may be, for example, a traditional type of keypad or keyboard, a mouse, a microphone for receiving voice signals, a camera, and various other types of input means for detecting or receiving various forms of user input. The output unit may be, for example, a display that outputs images, a speaker that outputs sound, a haptic device that generates vibrations, and various other forms of output means.

[0046] Additionally, at least some of the components within the electronic device (100) may be implemented as an integrated unit or as a single or multiple unit. At least some of the components within the electronic device (100) may be connected to each other via a bus, GPIO (general purpose input / output), SPI (serial peripheral interface) or MIPI (mobile industry processor interface), etc., to exchange data and / or signals.

[0047] A processor (110) is a component capable of performing operations or data processing regarding the control and / or communication of each component of an electronic device (100). For example, the processor (110) can control the electronic device (100) overall by executing programs stored in memory (120) within the electronic device (100). The processor (110) may be implemented as a CPU (central processing unit), GPU (graphics processing unit), AP (application processor), etc., provided within the electronic device (100), but is not limited thereto. Unless there are special circumstances, in this disclosure, the processor (110) may refer to a set of one or more processors (110).

[0048] The processor (110) may be implemented as a computer or a similar device according to hardware, software, or a combination thereof. In terms of hardware, the processor (110) may be implemented in the form of an electronic circuit that processes electrical signals to perform control functions, and in terms of software, it may be implemented in the form of a program that drives the hardware processor (110). Meanwhile, unless otherwise specifically mentioned in the description below, the operation of the electronic device (100) may be interpreted as being performed by the control of the processor (110). That is, when modules implemented in the electronic device (100) are executed to perform the image data processing method described below, the modules may be interpreted as controlling the processor (110) to perform the operations of the electronic device (100) below.

[0049] Memory (120) is hardware that stores various data processed within an electronic device (100), and memory (120) can temporarily or semi-permanently store data processed and data to be processed in the electronic device (100). For example, data regarding an operating system (OS) for operating the electronic device (100) can be stored in the memory (120) of the electronic device (100). Memory (120) may include random access memory (RAM) such as dynamic random access memory (DRAM) and static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), CD-ROM, Blu-ray or other optical disc storage, hard disk drive (HDD), solid state drive (SSD), or flash memory. Such memory (120) may be provided as an embedded type or a removable type. Additionally, memory (120) may store instructions for the operation of the processor (110). Unless there are special circumstances, in the present disclosure, memory (120) may mean a set of one or more memories (120).

[0050] The method for improving artificial intelligence accuracy of the present disclosure, performed by the electronic device (100) of FIG. 1, may also be implemented by a computer-readable, non-transient computer-readable storage medium (or non-transient recording medium) for the operation. The method for improving artificial intelligence accuracy may be implemented as a software module or algorithm and may be stored on a computer-readable recording medium as computer-readable code or program instructions executable on a processor (110). Here, computer-readable recording media include magnetic storage media (e.g., ROM (read-only memory), RAM (random-access memory), floppy disk, hard disk, etc.) and optical reading media (e.g., CD-ROM, DVD (Digital Versatile Disc)). Computer-readable recording media may be distributed across networked computer systems, so that computer-readable code can be stored and executed in a distributed manner. The medium is computer-readable, stored in memory (120), and can be executed on a processor (110).

[0051] An electronic device (100) according to one embodiment may further include a display (not shown). The display may visually provide various information to the outside under the control of a processor (110). The display may include a touch circuit configured to detect a user's touch or a sensor circuit configured to measure the intensity of the force generated by the touch.

[0052] The electronic device (100) of FIG. 1 can operate to design an AI based on a Monte-Carlo simulation algorithm based on the hardware configuration as described above, set a region of interest for executing a local Monte-Carlo simulation within a vulnerable area where the AI ​​is judged to be vulnerable among the application areas of the AI, execute a local Monte-Carlo simulation on the region of interest based on the AI, and verify information regarding the improved accuracy of the AI ​​based on the execution of the local Monte-Carlo simulation on the region of interest. Here, the AI ​​may be an AI linked to the electronic device (100) to provide computational results through the execution of a Monte-Carlo simulation, and the improved accuracy of the AI ​​may be determined based on the fact that the deep learning data obtained through the execution of a local Monte-Carlo simulation on the region of interest is additionally reflected in the deep learning data obtained through the execution of a Monte-Carlo simulation on the application area. In other words, the improved accuracy of the AI ​​in Fig. 1 can be determined based on the fact that the deep learning data obtained through the execution of a local Monte Carlo simulation of the region of interest is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation of the application region, and the case where it is not reflected.

[0053] AI based on Monte Carlo simulation algorithms faces the problem of performance limitations when dealing with a 'very large number of cases' due to the large scope of application of the AI. In response to this, the present disclosure identifies vulnerable areas within the scope of application of AI based on Monte Carlo simulation algorithms, and by setting a subset range of the Region of Interest (ROI) within the vulnerable areas, performs local Monte Carlo simulation on a scope limited to the ROI, thereby improving the accuracy of AI based on Monte Carlo simulation algorithms.

[0054] Regarding the execution of local Monte Carlo simulations, it is necessary to view the overall AI design from the perspective of focusing on and partially solving specific areas of interest—such as local life-and-death issues in Go or areas with frequent medical errors in diagnosis—much like the thinking of traditional human experts.

[0055] In one embodiment, since the Go AI is vulnerable to fighting in the central area during the opening phase, instead of setting it to simulate all possible cases in the entire 19x19 grid area of ​​the standard Go specifications from the initial AI design stage, the AI's accuracy can be increased by determining the vulnerable area, setting a subset of the area of ​​interest among the determined vulnerable areas, limiting the simulation area to that area of ​​interest, and intensively executing a local MCTS algorithm simulation within that limited range. For example, the local MCTS simulation can be intensively executed only within the central 10x10 grid area of ​​the entire 19x19 grid area.

[0056] Compared to the N*N line Go, the (N+2)*(N+2) line Go, which adds one line to each of the top, bottom, left, and right, adds 4N+4 stones, causing the total number of possible arrangements to increase exponentially compared to the existing N*N line Go. The 19*19 line Go, which is currently the most commonly used, can be considered very vast for human experts, or professional players, when considering the total number of possibilities. The search space for the 19*19 line is calculated to be approximately 10 to the power of 360, meaning that even existing high-performance supercomputers, which calculate all possible cases, are still woefully inadequate to compute all of them. This is evidenced by the fact that Go, a representative two-player zero-sum game with complete information determination, was the only board game until the 20th century in which a computer could not defeat a human world champion.

[0057] According to the principles of the MCTS algorithm, the probability that an AI possesses deep learning data regarding large-scale water battles that are concentrated in a limited area of ​​M*M lines (N > M) rather than the entire N*N line is a very rare probability in terms of information entropy. Therefore, deep learning data resulting from the execution of MCTS simulations on a limited area of ​​M*M lines can become a weakness that the AI ​​has not learned in advance. Conversely, if we utilize deep learning data on a limited area of ​​M*M lines (N > M) that is easy to obtain due to its relatively low computational cost, and manage the AI ​​to solve the given problem by reflecting and substituting this data into the deep learning data on the entire N*N line—much like the partial life-and-death thinking of human experts—then the accuracy and efficiency of the AI ​​can be improved.

[0058] In one embodiment, when the AI ​​that runs a simulation in conjunction with an electronic device (100) is a Go AI, an M*M line region of interest may be set within a central part of the entire N*N line Go board area where the AI ​​is judged to be weak (N > M). In this case, deep learning data obtained through the execution of a local Monte Carlo simulation on the M*M line region of interest may be additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation on the entire N*N line Go board area, thereby improving the accuracy of the AI. According to the embodiment, instead of utilizing only the deep learning data obtained through the execution of a Monte Carlo simulation on the entire N*N line Go board area as in the conventional method, deep learning data obtained through the execution of a local Monte Carlo simulation on the central M*M line area judged to be weak may also be additionally or separately utilized for the operation of the AI. The electronic device (100) can determine and verify the improved accuracy of the Go AI by comparing the case where deep learning data obtained through the execution of local Monte Carlo simulation as described above is reflected and the case where it is not.

[0059] In one embodiment, when the AI ​​that executes a simulation in conjunction with the electronic device (100) is a medical diagnostic AI, a medical diagnostic interest area may be established among the medical diagnostic areas where the rate of medical error judgment by the AI ​​is judged to be high. In this case, deep learning data obtained through the execution of a local Monte Carlo simulation on the medical diagnostic interest area may be additionally reflected in the deep learning data already obtained through the execution of a Monte Carlo simulation on the medical diagnostic area, thereby improving the accuracy of the AI. According to the embodiment, instead of utilizing only the deep learning data obtained through the execution of a Monte Carlo simulation on the entire medical diagnostic area as in the conventional method, deep learning data obtained through the execution of a local Monte Carlo simulation on the medical diagnostic interest area, which is judged to be a weakness due to a high rate of medical error judgment, may also be additionally or separately utilized for the operation of the AI. The electronic device (100) can judge and verify the improved accuracy of the medical diagnostic AI by comparing the case where the deep learning data obtained through the execution of the local Monte Carlo simulation as described above is reflected with the case where it is not.

[0060] Meanwhile, AI may be designed to allow for the application of alternative algorithms other than the Monte Carlo simulation algorithm. Although the Monte Carlo algorithm, used as a sampling method for conventional machine learning, theoretically possesses randomness and can achieve homogeneity as the number of samples increases, classical Monte Carlo simulation operates on the principle of calculating the ratio of points within a target area to the total area by randomly selecting and plotting points. This presents a problem in that it is somewhat unsuitable for AI systems that prioritize efficiency. Originally, the Monte Carlo method was an idea adopted by a mathematician named Ulam during the manufacturing process of the hydrogen bomb; due to its characteristics derived from physical phenomena, randomness could be reflected, but in machine learning, such randomness can lead to inefficiency. The ultimate goal of randomization in Monte Carlo algorithms is to achieve homogenization; however, strictly speaking, true randomness does not practically exist in Monte Carlo algorithms, and they possess a weaker form of pseudo-randomness. Furthermore, due to inaccuracy, there is an inefficient aspect in that many attempts and resources are required based on the law of large numbers. While pure randomness could be meaningful as a sample, accurately estimating a ratio corresponding to the law of large numbers requires many attempts. Moreover, since it may require not just simply plotting points but also implementing detailed rules that follow from those points, a significant amount of time and resources are required. Accordingly, instead of the classical Monte Carlo algorithm, one may consider applying an improved algorithm that acknowledges the limitations of randomness and focuses on homogenization—the ultimate goal of randomization—similarly to the present disclosure, such as the Latin Hypercube sampling method, which is a ratio estimation technique in the form of appropriate lattice points and grid cells that is easy to implement and guarantees uniformity.

[0061] Since grid points are determined by simpler rules and ensuring uniformity is more important than randomness, this may constitute a new approach distinct from the Monte Carlo algorithm, and the estimation of associated values ​​can also be more accurate and efficient. The application of the Latin hypercube sampling method can be understood as being based on the idea that, when calculating area ratios, ensuring uniformity by finding the number of minimum component members on a coordinate plane such as a chessboard is actually the most efficient method. Compared to general Monte Carlo simulations, the Latin hypercube sampling method is known to have the advantage of a homogeneous sample distribution and the ability to pursue the same level of statistical accuracy with a small number of samples; although it is even included as a menu item in existing statistical programs, there have been no cases of it being applied to improve the accuracy of AI, so the embodiment according to the present disclosure can be interpreted as a new attempt.

[0062] In one embodiment, when the AI ​​applies a Monte Carlo simulation algorithm, it may be configured to apply a Latin hypercube algorithm instead of the Monte Carlo simulation algorithm depending on conditions related to uniformity and / or efficiency. Specifically, when the AI ​​applies a Monte Carlo simulation algorithm, if the uniformity confirmed by detailed areas regarding the AI's computation and results targeting the application area, vulnerable area, and / or area of ​​interest is below a certain level, or if the computational efficiency of the AI ​​targeting the application area, vulnerable area, and / or area of ​​interest is determined to be below a certain level, the AI ​​may apply a Latin hypercube algorithm instead of the Monte Carlo simulation algorithm. In this case, the AI ​​may perform operations according to various embodiments of the present disclosure described above based on the Latin hypercube algorithm, and the electronic device (100) may also verify information regarding the improved accuracy of the AI ​​based on the Latin hypercube algorithm.

[0063] In one embodiment, to fundamentally prevent frequent errors of the AI ​​occurring in a specific area of ​​the AI's application domain, if the AI's inference value is determined to correspond to a state such as a self-evident bug or an incorrect inference result that deviates from regulations, the AI ​​may be forced to ignore the inference process and result according to the specific area and determine the inference process and result through a self-algorithm pre-configured in the AI ​​to prevent errors, thereby further increasing the accuracy of the AI ​​based on the Monte Carlo simulation algorithm. In other words, if the AI ​​is determined to correspond to an application error state or an inference error state regarding a specific area of ​​the AI's application domain, the AI ​​may be configured to exclude the application result for that specific area based on a processing algorithm pre-configured in the AI ​​and obtain only the application result for the remaining area excluding the specific area.

[0064] In one embodiment, when the electronic device (100) determines and confirms the enhanced accuracy based on a local Monte Carlo simulation of the AI ​​for a region of interest, the electronic device (100) may store information by pairing the size of the region of interest among vulnerable regions with the enhanced accuracy of the AI ​​when a local Monte Carlo simulation is performed on the region of interest of that size. Additionally, the electronic device (100) may change the size of the region of interest in stages according to a certain degree within the vulnerable region, and store information as described above for each region of interest and the enhanced accuracy of the AI ​​corresponding to each region of interest, and based on this, determine the size of the region of interest within the vulnerable region where the accuracy is most enhanced.

[0065] Taking Go AI as an example, when the area where the AI ​​is weak within the entire standard 19x19 line Go board area is defined as the central part of the 10x10 line area, the electronic device (100) stores: 1) information including a pair of a 5x5 line area of ​​interest and the improved accuracy of the AI ​​confirmed when deep learning data obtained by performing a local Monte Carlo simulation on the 5x5 line area of ​​interest is reflected; 2) information including a pair of a 6x6 line area of ​​interest and the improved accuracy of the AI ​​confirmed when deep learning data obtained by performing a local Monte Carlo simulation on the 6x6 line area of ​​interest is reflected; ..., 3) information including a pair of a 10x10 line area of ​​interest and the improved accuracy of the AI ​​confirmed when deep learning data obtained by performing a local Monte Carlo simulation on the 10x10 line area of ​​interest is reflected; and based on the stored information, the 5x5 line, 6x5 line, ..., Among the 10x10 lines, the size of the optimal region of interest with the highest enhanced accuracy of the AI ​​can be determined. Subsequently, when the electronic device (100) needs to perform a Monte Carlo simulation on the entire 19x19 line chessboard area in other similar situations, the performance of the AI ​​can be improved by performing a local Monte Carlo simulation based on the previously determined optimal region of interest and managing the deep learning data obtained through this so that it is reflected in the operation of the AI.

[0066] In summary, the electronic device (100) can set multiple regions of interest of different sizes within a vulnerable area, and can verify the improved accuracy of the AI ​​determined by each region of interest in which the acquired deep learning data is reflected by performing a local Monte Carlo simulation for each of the multiple regions of interest, and can identify the region of interest of the optimal size where the accuracy is most improved. In this way, the electronic device (100) identifies the region of interest of the optimal size within the vulnerable area in terms of accuracy, and can utilize the associated information for AI operation in similar situations in the future.

[0067] Meanwhile, depending on the size of the vulnerable area, there may be a significant difference in the size of the region of interest that can be set. Taking the aforementioned Go AI as an example, from the perspective of AI simulation, it can be understood that there is a considerable difference in the size of the region of interest when it is set to 5x5 lines versus 10x10 lines. Even if the AI ​​can achieve improved accuracy when the region of interest is set to 10x10 lines compared to 5x5 lines when performing local Monte Carlo simulations, the 10x10 line region of interest generates a much larger computational load than the 5x5 line region of interest. Therefore, in terms of AI performance, where computational load is considered in addition to accuracy, it may be difficult to conclude that the 10x10 line region of interest is the optimal size of the region of interest more appropriate than the 5x5 line region of interest.

[0068] Therefore, in determining the optimal size region of interest for the AI ​​to execute local Monte Carlo simulations, not only the enhanced accuracy identified corresponding to the size of each region of interest as described above, but also the computational load identified corresponding to the size of each region of interest can be reflected. That is, in determining the optimal size region of interest for the AI's performance among regions of interest that can be set to various sizes within the vulnerable region, the electronic device (100) can determine the optimal size region of interest by considering the enhanced accuracy identified for each region of interest and the degree of computational load for each region of interest.

[0069] First, the electronic device (100) can set and display each enhanced accuracy confirmed in correspondence with the size of each region of interest as a percentage value. Additionally, the electronic device (100) can determine the degree of each computational load confirmed for each region of interest of each size as a percentage value between 0% and 100% by assuming that the case where there is no computational load is 0% and the case where the AI ​​cannot operate due to the computational load is 100%. Based on this, the electronic device (100) can determine the optimal size of the region of interest by combining the percentage value of each enhanced accuracy confirmed in correspondence with the size of each region of interest and the percentage value of each degree of computational load confirmed in correspondence with the size of each region of interest. Since a higher percentage value of the computational load indicates a greater computational load that hinders the performance of the AI, it can be set to be reflected as a negative (-) value when determining the optimal size of the region of interest. Furthermore, the electronic device (100) can consider whether to focus on accuracy or prevent computational load in the performance of the AI, and can dynamically set an accuracy weight for enhanced accuracy and a computational load weight for the degree of computational load, thereby allowing the optimal size of the region of interest to be determined differently depending on the weight.

[0070] The above example for determining the optimal size of the region of interest can be summarized as a mathematical formula as [Equation 1] below.

[0071] [Mathematical Formula 1]

[0072] argmax((f(x))

[0073] f(x) = *Improved Accuracy (%) - *Computational Load Level (%)

[0074] (x = size of each region of interest, = Accuracy weight, = Computational load weight)

[0075] According to [Equation 1], the degree of improved accuracy and computational load for each region of interest can be checked, and the value of f(x) can be determined by multiplying each weight, and the size of the region of interest with the largest f(x) value can be determined. In terms of accuracy and computational load, the electronic device (100) identifies the region of interest of the size determined by [Equation 1] as the region of interest of the optimal size within the vulnerable region, and the associated information can be utilized for AI operation in similar situations in the future.

[0076] It is obvious that each piece of information or data generated during the process of the electronic device (100) performing a method to improve artificial intelligence accuracy can be combined in various forms.

[0077] As examined above, existing AI based on Monte Carlo simulation algorithms inherently had a vulnerability due to the large number of cases for the target service. However, according to the present disclosure, instead of simulating all possible cases for the entire system, the simulation area is reduced to the vulnerable region. In particular, by setting a subset of the region of interest and executing a local Monte Carlo simulation algorithm within that limited range, the accuracy of the AI ​​can be increased.

[0078] In the foregoing, although all components constituting the embodiments disclosed in this document have been described as being combined or operating in combination, the embodiments disclosed in this document are not necessarily limited to such embodiments. That is, within the scope of the purposes of the embodiments disclosed in this document, all components may be selectively combined in one or more ways to operate.

[0079] Furthermore, terms such as "include," "compose," or "have" as described above, unless specifically stated otherwise, mean that the relevant component may be inherent; thus, they should be interpreted as allowing for the inclusion of additional components rather than excluding them. All terms, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the embodiments disclosed in this document pertain, unless otherwise defined. Commonly used terms, such as those defined in advance, should be interpreted in accordance with their meaning in the context of the relevant technology and, unless explicitly defined in this document, should not be interpreted in an ideal or overly formal sense.

[0080] The foregoing description is merely an illustrative explanation of the technical concept disclosed in this document, and a person skilled in the art to which the embodiments disclosed in this document pertain can make various modifications and variations within the scope of the essential characteristics of the embodiments disclosed in this document. Accordingly, the embodiments disclosed in this document are intended to explain, not limit, the technical concept of the embodiments disclosed in this document, and the scope of the technical concept disclosed in this document is not limited by these embodiments. The scope of protection of the technical concept disclosed in this document shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of this document.

Claims

Claim 1 A method for improving the accuracy of Artificial Intelligence (AI) of an electronic device comprises: a step of designing an AI based on a Monte-Carlo simulation algorithm; a step of setting a region of interest for executing a local Monte-Carlo simulation within a vulnerable region where the AI ​​is determined to be vulnerable among the application regions of the AI; a step of executing a local Monte-Carlo simulation for the region of interest based on the AI; and a step of confirming information regarding the improved accuracy of the AI ​​based on the execution of the local Monte-Carlo simulation for the region of interest, wherein the improved accuracy is determined based on the fact that deep learning data obtained through the execution of the local Monte-Carlo simulation for the region of interest is additionally reflected in deep learning data obtained through the execution of the Monte-Carlo simulation for the application region, and the AI ​​accuracy improvement method comprises: a step of setting a plurality of regions of interest having different sizes within the vulnerable region; and a step of confirming each improved accuracy determined for the AI ​​based on the execution of a local Monte-Carlo simulation for each of the regions of interest included in the plurality of regions of interest. A method for improving AI accuracy, further comprising the step of setting information including the size of each of the regions of interest and each of the enhanced accuracy. Claim 2 A method for improving AI accuracy according to claim 1, wherein the AI ​​is configured to allow the application of a Latin Hypercube algorithm instead of the Monte Carlo simulation algorithm based on conditions related to uniformity and efficiency. Claim 3 In claim 1, when the AI ​​corresponds to a Go AI, an M*M line region of interest is set within a central part area where the AI ​​is judged to be weak among the entire N*N line Go board area, and the enhanced accuracy is determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation on the M*M line region of interest is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation on the entire N*N line Go board area, and M is smaller than N, a method for improving AI accuracy. Claim 4 A method for improving AI accuracy according to claim 1, wherein when the AI ​​corresponds to a medical diagnostic AI, a medical diagnostic interest region is set among the medical diagnostic regions in which the medical error judgment rate by the AI ​​is determined to be high, and the improved accuracy is determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation for the medical diagnostic interest region is additionally reflected in deep learning data obtained through the execution of a Monte Carlo simulation for the medical diagnostic region. Claim 5 A method for improving AI accuracy according to claim 1, wherein if it is determined that the AI ​​corresponds to an application error state with respect to a specific area among the application areas, the application result of the AI ​​for the specific area is excluded based on a processing algorithm pre-set in the AI, and only the application result of the AI ​​for the remaining area excluding the specific area among the application areas is obtained. Claim 6 delete Claim 7 The AI ​​accuracy enhancement method according to claim 1 further comprises: a step of determining the degree of computational load determined when executing a local Monte Carlo simulation for each of the plurality of regions of interest included in the plurality of regions of interest; and a step of determining a region of interest of an optimal size for the performance of the AI ​​among the plurality of regions of interest based on the respective enhanced accuracy and the respective degree of computational load. Claim 8 An electronic device for performing a method to improve the accuracy of Artificial Intelligence (AI) comprises: a processor; and one or more memories for storing one or more instructions, wherein, when the one or more instructions are executed, the processor comprises: a step of designing an AI based on a Monte-Carlo simulation algorithm; a step of setting a region of interest for executing a local Monte-Carlo simulation within a vulnerable region where the AI ​​is determined to be vulnerable among the application areas of the AI; and a step of executing a local Monte-Carlo simulation for the region of interest based on the AI. An electronic device that controls the processor to perform the step of verifying information regarding the enhanced accuracy of the AI ​​based on the execution of a local Monte Carlo simulation for the region of interest, wherein the enhanced accuracy is determined based on the fact that the deep learning data obtained through the execution of a local Monte Carlo simulation for the region of interest is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation for the application area, and the processor further performs the steps of: setting a plurality of regions of interest having different sizes within the vulnerable area; verifying each enhanced accuracy determined for the AI ​​based on the execution of a local Monte Carlo simulation for each of the regions of interest included in the plurality of regions of interest; and setting information including the size of each region of interest and each enhanced accuracy. Claim 9 In claim 8, the electronic device wherein the AI ​​is configured to enable the application of a Latin Hypercube algorithm instead of the Monte Carlo simulation algorithm based on conditions related to uniformity and efficiency. Claim 10 In claim 8, when the AI ​​corresponds to a Go AI, an M*M line region of interest is set within a central portion of the entire N*N line Go board area where the AI ​​is judged to be weak, and the enhanced accuracy is determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation on the M*M line region of interest is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation on the entire N*N line Go board area, and M is smaller than N, an electronic device. Claim 11 An electronic device according to claim 8, wherein, when the AI ​​corresponds to a medical diagnostic AI, a medical diagnostic interest region is set in which the medical error judgment rate by the AI ​​is determined to be high among the medical diagnostic regions, and the enhanced accuracy is determined based on the fact that deep learning data obtained through the execution of a local Monte Carlo simulation of the medical diagnostic interest region is additionally reflected in the deep learning data obtained through the execution of a Monte Carlo simulation of the medical diagnostic region. Claim 12 An electronic device according to claim 8, wherein, when it is determined that the AI ​​corresponds to an application error state with respect to a specific area among the application areas, the application result of the AI ​​for the specific area is excluded based on a processing algorithm pre-set in the AI, and only the application result of the AI ​​for the remaining area excluding the specific area among the application areas is obtained. Claim 13 delete Claim 14 An electronic device according to claim 8, wherein the processor further performs the steps of: determining the degree of computational load for each region of interest included in the plurality of regions of interest when executing a local Monte Carlo simulation; and determining the region of interest of an optimal size for the performance of the AI ​​among the plurality of regions of interest based on the respective enhanced accuracy and the respective degree of computational load.

Images