Resource combination optimization method and device based on quantum network, equipment and medium

Abstract

The invention relates to the technical field of intelligent decision making, can be applied to business system platforms of financial science and technology, medical treatment and health and the like, and discloses a resource combination optimization method, device, equipment and medium based on a quantum network, and the method comprises the steps: obtaining candidate resource combinations in historical transaction data, and analyzing a target analysis value corresponding to each candidate resource combination; constructing a supervised learning sample pair according to the candidate resource combination and the target analysis value, and constructing a quantum neural network model of the candidate resource combination by using the supervised learning sample pair; performing form optimization on a quantum function of the quantum neural network model to obtain a standard quadratic form; generating a quantum optimization target of the candidate resource combination according to the standard quadratic form, and analyzing an expected energy parameter corresponding to the quantum optimization target by using a quantum optimization algorithm; and analyzing a target quantum state of the expected energy parameter, and optimizing the candidate resource combination according to the target quantum state to obtain a target resource combination. And the accuracy of resource combination optimization is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent decision-making, and in particular to a resource combination optimization method and device based on a quantum network, equipment and a medium. BACKGROUND

[0002] In recent years, combination optimization is the core means for solving resource allocation and decision-making problems in multiple fields. With the expansion of data scale and the increase of problem complexity, traditional combination optimization methods gradually expose limitations, and quantum computing provides a new technical path for solving high-dimensional and nonlinear combination optimization problems due to its parallelism and superposition characteristics. However, the existing technology still has significant defects in practical application.

[0003] In the field of medical health, treatment scheme combination optimization (such as drug combination and therapy matching) is an important way to improve treatment effect and reduce medical risk, which needs to consider side effects and treatment costs while ensuring treatment effect. However, existing methods mostly use simple linear indicators (such as treatment effect-risk difference), which are difficult to capture the nonlinear and non-convex relationship between the two, resulting in insufficient accuracy of combination scheme evaluation.

[0004] In the field of financial technology business, resource combination optimization is the key to efficient allocation of funds, and its goal is to select the optimal resource allocation scheme for investors while balancing returns and risks. However, relying solely on historical backtesting data to select the optimal combination cannot be generalized to combinations that have not appeared before, resulting in low accuracy of combination schemes.

[0005] Current quantum portfolio optimization research mostly focuses on accelerating the solution of combination selection problems, and does not fully utilize the potential of quantum computing in modeling complex nonlinear functions. In particular, in the modeling of the non-convex and nonlinear target of Sharpe ratio, traditional statistical backtesting or black box estimation methods are still relied on, resulting in low accuracy in resource combination optimization. SUMMARY The present application provides a resource combination optimization method and device based on a quantum network to solve the technical problem of low accuracy in resource combination optimization.

[0006] In a first aspect, a resource combination optimization method based on a quantum network is provided, comprising: Obtaining candidate resource combinations in preset historical transaction data, and analyzing target analysis values corresponding to each candidate resource combination; Constructing a supervised learning sample pair according to the candidate resource combinations and the target analysis values, and constructing a quantum neural network model of the candidate resource combinations using the supervised learning sample pair; Formally optimizing a quantum function of the quantum neural network model to obtain a standard quadratic form; generate a quantum optimization objective of the candidate resource combination according to the standard quadratic form, and analyze an expected energy parameter corresponding to the quantum optimization objective by using a preset quantum optimization algorithm; analyze a target quantum state of the expected energy parameter, optimize the candidate resource combination according to the target quantum state, and obtain a target resource combination.

[0007] In a second aspect, a resource combination optimization apparatus based on a quantum network is provided, and the apparatus comprises: a target analysis value analysis module configured to obtain candidate resource combinations in preset historical transaction data, and analyze target analysis values corresponding to each candidate resource combination; a quantum neural network model construction module configured to construct a supervised learning sample pair according to the candidate resource combinations and the target analysis values, and construct a quantum neural network model of the candidate resource combinations by using the supervised learning sample pair; a quantum function form optimization module configured to perform form optimization on a quantum function of the quantum neural network model to obtain a standard quadratic form; an expected energy parameter analysis module configured to generate a quantum optimization objective of the candidate resource combination according to the standard quadratic form, and analyze an expected energy parameter corresponding to the quantum optimization objective by using a preset quantum optimization algorithm; a candidate resource combination optimization module configured to analyze a target quantum state of the expected energy parameter, optimize the candidate resource combination according to the target quantum state, and obtain a target resource combination.

[0008] In a third aspect, a computer device is provided, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above-mentioned resource combination optimization method based on a quantum network when executing the computer program.

[0009] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program implements the steps of the above-mentioned resource combination optimization method based on a quantum network when executed by a processor.

[0010] In the scheme implemented by the resource combination optimization method, device, equipment and medium based on a quantum network, a candidate resource combination in preset historical transaction data can be acquired by a client, and a target analysis value corresponding to each candidate resource combination can be analyzed. A quantum neural network model of the candidate resource combination can be constructed by using a supervised learning sample pair constructed according to the candidate resource combination and the target analysis value. A quantum function of the quantum neural network model can be formally optimized to obtain a standard quadratic form. A quantum optimization target of the candidate resource combination can be generated according to the standard quadratic form, and an expected energy parameter corresponding to the quantum optimization target can be analyzed by using a preset quantum optimization algorithm. A target quantum state of the expected energy parameter can be analyzed, the candidate resource combination can be optimized according to the target quantum state, and a target resource combination can be obtained. The target resource combination is fed back to the client. In the present application, an optimization framework with a target analysis value prediction function as the core is constructed. The quantum neural network is used for modeling in view of the characteristics of the target analysis value, such as nonlinearity, non-convexity and non-differentiability. The complex non-convex and nonlinear prediction target function is converted into a standard format by using a quadratic polynomial regression method, so that the target analysis value in the form of a ratio can be embedded in the structure of the quantum approximation optimization algorithm, a bridge between high-dimensional nonlinear modeling and quantum solving is built, and the realizability and operation efficiency of the overall system are improved. The quantum parallel search and quantum entanglement of the combination space are realized by combining the approximate optimization algorithm, and the optimal solution in the high-dimensional discrete space is mined. The potential of quantum computing in financial intelligent decision-making is embodied, and the technical problem of low accuracy in resource combination optimization is solved. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0012] Figure 1 is an application environment schematic diagram of the resource combination optimization method based on a quantum network in an embodiment of the present application; Figure 2 is a flowchart of the resource combination optimization method based on a quantum network in an embodiment of the present application; Figure 3 is Figure 2 is a specific implementation flowchart of step S3 in Figure 4 is a specific implementation flowchart of step S4 in Figure 2 Figure 5 ​is a structural schematic diagram of a quantum network-based resource combination optimization device in an embodiment of the present application; Figure 6 is a structural schematic diagram of a computer device in an embodiment of the present application; Figure 7 is another structural schematic diagram of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0013] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0014] The quantum network-based resource combination optimization method provided by the embodiments of the present application can be applied to, for example, Figure 1In this application environment, the client communicates with the server via a network. The server can obtain candidate resource combinations from preset historical transaction data through the client, analyze the target analysis value corresponding to each candidate resource combination; construct supervised learning sample pairs based on the candidate resource combinations and the target analysis values, and use the supervised learning sample pairs to construct a quantum neural network model of the candidate resource combinations; perform formal optimization on the quantum function of the quantum neural network model to obtain the standard quadratic form; generate the quantum optimization target of the candidate resource combinations based on the standard quadratic form, and analyze the expected energy parameter corresponding to the quantum optimization target using a preset quantum optimization algorithm; analyze the target quantum state of the expected energy parameter, optimize the candidate resource combinations based on the target quantum state to obtain the target resource combination, and feed the target resource combination back to the client. In this invention, a quantum optimization algorithm with the target analysis value prediction function as the core is constructed. This invention employs a quantum neural network to model the nonlinear, nonconvex, and non-differentiable characteristics of the target analysis value. A quadratic Boolean regression method is used to transform the complex nonlinear prediction objective function into a standard format, allowing the ratio-based target analysis value to be embedded in the structure of a quantum approximation optimization algorithm. This bridges the gap between high-dimensional nonlinear modeling and quantum solution, improving the overall system's feasibility and computational efficiency. The approximation optimization algorithm enables quantum parallel search and quantum entanglement in combinatorial space, achieving quasi-exponential speedup compared to traditional heuristic algorithms. It is particularly suitable for optimal solution mining in high-dimensional discrete spaces, demonstrating the potential of quantum computing in financial intelligent decision-making and addressing the technical problem of low accuracy in resource combinatorial optimization. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster. Specific embodiments are described in detail below.

[0015] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a resource combinatorial optimization method based on quantum networks provided in an embodiment of the present invention includes the following steps: S1. Obtain candidate resource combinations from preset historical transaction data and analyze the target analysis value corresponding to each candidate resource combination.

[0016] In this embodiment of the invention, the candidate resource combination is a possible investment portfolio composed of multiple resources selected from preset historical transaction data.

[0017] In detail, each candidate resource combination is represented by a binary vector of length n. ,in, Indicates selecting the first Each resource (also referring to assets in an investment) enters the portfolio. denotes no selection, each element in the vector takes a value of 0 or 1, and a value of 1 indicates that the resource corresponding to the position is included in the combination, and a value of 0 indicates that the resource is not included in the combination. For example, when n = 5, a certain candidate resource combination can be represented as [1, 0, 1, 0, 1], that is, the first, third and fifth resources are included, and the second and fourth resources are not included.

[0018] Further, for each candidate resource combination , the target analysis value in the historical interval, that is, the backtest Sharpe ratio is calculated.

[0019] In the embodiment of the present application, the target analysis value is the backtest Sharpe ratio, that is, the ratio of the expected value to the maximum amplitude value.

[0020] In the embodiment of the present application, the analysis of the target analysis value corresponding to each candidate resource combination comprises: obtaining the historical resource data interval corresponding to each candidate resource combination; extracting the expected value and the maximum amplitude value corresponding to each candidate resource combination according to the historical resource data interval within a preset time period; calculating the target analysis value corresponding to each candidate resource combination according to the expected value and the maximum amplitude value.

[0021] In detail, the historical resource data interval refers to the range covered by the trading data of all resources in the candidate resource combination within a certain continuous time period in the past. For example, if the investment performance in the past 3 years is analyzed, the historical resource data interval can be set as January 2021 to December 2023, covering the daily closing price, yield and other data of each resource within the interval; the preset time period can be set according to the investment strategy, such as monthly, quarterly or annually.

[0022] Specifically, the expected value refers to the average yield of the candidate resource combination within the preset time period, and the calculation method is the average value of the weighted sum of the yields of the resources in the combination according to the inclusion weight (the weight corresponding to the resources accounted for by the candidate resource combination is 0 or 1). For example, a certain candidate resource combination includes 2 resources, and the monthly average yield of resource A is 1% and the monthly average yield of resource B is 0.8% in the past 12 months, then the expected value of the combination is 0.9%. The maximum amplitude value, that is, the maximum drawdown, refers to the maximum proportion of the combination net value falling from the historical peak value to the valley value within the preset time period, and the calculation method is (peak net value-valley net value) / peak net value. For example, the combination net value falls from 10 yuan to 7 yuan and then rises, then the maximum amplitude value is 30%, and the target analysis value is the Sharpe ratio, and the calculation method is the ratio of the expected value to the maximum amplitude value. Following the above example, if the expected value of the combination is 0.9% and the maximum amplitude value is 30%, then the target analysis value is 0.03.

[0023] Exemplarily, in a financial scenario, preset historical transaction data is stock transaction records of a fund in the past 5 years, 100 candidate resource combinations are collected from the historical transaction data, each combination is composed of 5 stocks (represented by a binary vector, for example, [1, 0, 1, 0, 1] represents that the first, third and fifth stocks are included), when analyzing the target analysis value, a preset time period is 1 year, a historical resource data interval of each combination is extracted, an expected value and a maximum amplitude value of the combination in the interval are calculated, and thus the target analysis value is determined.

[0024] In addition, in a medical scenario, preset historical data is tumor treatment scheme records of a hospital in the past 3 years, candidate resource combinations are different chemotherapy drug combinations (for example, a combination of 5 chemotherapy drugs, [1, 0, 1, 0, 1] represents that the first, third and fifth drugs are used), when analyzing the target analysis value, a preset time period is 6 months, a historical data interval of each drug combination is extracted, an expected value and a maximum amplitude value of the combination in the interval are calculated, and thus the target analysis value is determined.

[0025] Further, in addition to considering the limitation of the generalization ability of the historical data, the historical back test produces discrete label values, lacks a continuous derivable structure form, and is difficult to be converted into a target function that can be processed by a quantum optimizer, therefore, it is necessary to construct a quantum neural network model to process the target function.

[0026] S2, constructing a supervised learning sample pair according to the candidate resource combination and the target analysis value, and constructing a quantum neural network model of the candidate resource combination by using the supervised learning sample pair.

[0027] In the embodiment of the application, the supervised learning sample pair is a pair of data composed of a candidate resource combination and a target analysis value corresponding to the combination, that is, .

[0028] In detail, if the candidate resource combination is [1, 0, 1, 0, 1] and the corresponding target analysis value is 0.03, the supervised learning sample pair is ([1, 0, 1, 0, 1], 0.03); a plurality of such sample pairs constitute a sample set for training the model.

[0029] In the embodiment of the application, the quantum neural network model refers to a nonlinear mapping relationship fitting the combination and the target analysis value .

[0030] In the embodiment of the application, the quantum neural network model of the candidate resource combination is constructed by using the supervised learning sample pair, comprising: mapping the candidate resource combination in the supervised learning sample to a quantum state by using a preset angular encoding; The quantum state is nonlinearly mapped with a target analysis value in the supervised learning sample to obtain an initial quantum network model; The initial quantum network model is used to calculate the target expected value of a pre-acquired quantum state to be analyzed, and the real expected value corresponding to the quantum state to be analyzed is analyzed; The loss value between the target expected value and the real expected value is calculated, and the circuit parameters in the initial quantum network model are optimized according to the loss value; When the loss value is less than a preset loss threshold, the optimized initial quantum network model is used as the quantum neural network model of the candidate resource combination.

[0031] In detail, the input is mapped to a quantum state using angle encoding, and the number of quantum bits is the dimension of the input . The calculation method of angle encoding is to construct a mapping , which maps the input to the parameters of the rotation gate RY, which is used to rotate the state of the corresponding quantum bit, as follows: The initial quantum network model is composed of multiple layers of parameterized quantum gates, including single-qubit rotation gates (RX, RY, RZ) and two-qubit entanglement gates (CNOT). For example, for a candidate resource combination of n=2, a 4-layer quantum circuit can be designed, each layer containing a RY rotation gate and a CNOT gate. The nonlinear relationship between the candidate resource combination and the target analysis value is captured through the superposition and entanglement of quantum states.

[0032] Specifically, the quantum state to be analyzed refers to the quantum state of the candidate resource combination after angle encoding that has not participated in training; the target expected value is the output prediction of the initial quantum network model for the quantum state (i.e., the predicted Kappa ratio); the output quantum state is measured, and the expected value is extracted as the model output, based on the parameterized circuit The Pauli-Z operator is used to measure the projection of the quantum state on the Z-axis direction of the Bloch sphere, and its expected value range is , as follows: ; the real expected value is the actual Kappa ratio corresponding to the quantum state to be analyzed, for example, the real expected value of a candidate resource combination corresponding to a quantum state to be analyzed is 0.04, and the target expected value predicted by the initial model is 0.035.

[0033] Further, the loss value is calculated by using mean square error (MSE), and a classical optimizer (such as Adam) is used to update the angle parameters of the rotation gate in the quantum circuit to minimize the loss value, and when the loss value is less than a preset loss threshold, the initial quantum network model after optimization is taken as the quantum neural network model of the candidate resource combination: the preset loss threshold can be set according to the accuracy requirement, for example, 0.0001, if the loss value is reduced to 0.00008 (less than 0.0001) after multiple iterations, the optimization is stopped, and the model at this time is the quantum neural network model that can accurately predict the target analysis value of the candidate resource combination.

[0034] Exemplarily, in a financial scenario, the supervised learning sample pair is (candidate stock combination, Sharpe ratio), and when constructing the quantum neural network, the combination is mapped to a quantum state by using angle encoding: the first, third and fifth stocks correspond to quantum bits that are rotated by π / 2 through an RY gate, and the second and fourth stocks correspond to quantum bits that are rotated by 0. A 4-layer parameterized quantum circuit (containing an RY rotation gate and a CNOT entanglement gate) is designed to predict the target expected value (predicted Sharpe ratio) of the combination that does not participate in training, and the circuit parameters are adjusted through an Adam optimizer, and when the loss value is reduced to 0.0001 (less than the preset threshold 0.0005), the model is determined as the final quantum neural network model.

[0035] In addition, in a medical scenario, the supervised learning sample pair is (candidate drug combination, treatment benefit-risk ratio). When constructing the quantum neural network, the drug combination is mapped to a quantum state by using angle encoding: the quantum bits corresponding to the drugs used are rotated by π / 2, and the quantum bits corresponding to the drugs not used are rotated by 0. A 3-layer parameterized quantum circuit is designed to predict the target expected value (predicted benefit-risk ratio) of the new drug combination, and after optimizing the parameters, when the loss value is less than 0.05, the model is determined as the final quantum neural network model.

[0036] Further, through the superposition and entanglement characteristics of the quantum neural network, the problem that the traditional model is difficult to capture the nonlinear and non-convex relationship of the Sharpe ratio is solved, and the generalization prediction ability for unknown combinations is improved.

[0037] S3, formally optimizing the quantum function of the quantum neural network model to obtain a standard quadratic form.

[0038] In the embodiment of the application, the standard quadratic form refers to converting the quantum function of the quantum neural network model into an optimizable form, approximating the output result by using a quadratic polynomial regression method, and fitting it into a standard quadratic form.

[0039] In the embodiment of the application, referring to Figure 3 , the formal optimization of the quantum function of the quantum neural network model to obtain a standard quadratic form comprises: S31. Using the preset symmetric term quadratic coefficient matrix and linear term coefficient vector as a reference, fit the parameter values ​​in the quantum function to obtain the fitting result; S32. Analyze the fitting error corresponding to the fitting result based on the preset regression target; S33. Correct the fitting result based on the fitting error to obtain the standard quadratic form.

[0040] In detail, for the purpose of Transforming it into an optimizable form, and approximating its output using quadratic Boolean regression, we fit it to the following standard quadratic form: ,in: It is a symmetric quadratic coefficient matrix; This is the vector of coefficients for the first-order terms; This is a constant term (which can be ignored in optimization).

[0041] Specifically, the regression objective is to minimize the fitting error on the training samples: For example, if the sum of errors for 100 samples is 0.05, it indicates that the current fitting effect of Q and c needs to be improved. If the fitting error is large, adjustments are needed. and The element values ​​are refitted until the error meets the preset requirements (e.g., the total error is less than 0.01). The corrected standard quadratic form is: (The constant term b can be ignored and does not affect the optimization). After the fitting is completed, the Kalmar ratio prediction function can be embedded into the structure of the QUBO model.

[0042] Furthermore, by converting nonlinear quantum functions into standard quadratic forms that can be handled by quantum optimization algorithms (such as QAOA), the problem of quantum functions being difficult to use directly for quantum optimization is solved, thus bridging the gap for subsequent quantum solutions.

[0043] S4. Generate the quantum optimization objective of the candidate resource combination according to the standard quadratic form, and analyze the expected energy parameter corresponding to the quantum optimization objective using a preset quantum optimization algorithm.

[0044] In this embodiment of the invention, the quantum optimization objective refers to the objective function used to solve the quantum optimization algorithm. Its core is to maximize the objective analysis value (Karma ratio) while taking into account practical constraints (such as investment amount and handling fees).

[0045] In this embodiment of the invention, reference is made to Figure 4 As shown, the quantum optimization objective for generating the candidate resource combination according to the standard quadratic form includes: S41, obtain a combination constraint parameter of the candidate resource combination, and convert the combination constraint parameter into a penalty term parameter of the candidate resource combination; S42, obtain an optimization cost control parameter of an initial optimization target corresponding to the candidate resource combination; S43, combine the penalty term parameter, the optimization cost control parameter and the standard quadratic form to obtain a quantum optimization target of the candidate resource combination.

[0046] In detail, the combination constraint parameter refers to a limitation condition of the candidate resource combination, such as a total investment amount D (that is, the number of resources included in the combination is D). The penalty term parameter M is a positive number for strengthening the constraint, and the penalty term is When , the penalty term increases, so that the combination is excluded by the optimization algorithm). The optimization cost control parameter refers to a transaction fee rate T for measuring the cost of adjusting the portfolio, and the optimization calculation mode is When is the initial position, and is the current combination, and represents adjusting the portfolio, and 0 represents not adjusting the portfolio), the penalty term parameter, the optimization cost control parameter and the standard quadratic form are combined to obtain a quantum optimization target of the candidate resource combination.

[0047] Specifically, in order to use the quantum approximate optimization algorithm (QAOA) for solving, the above quadratic function is constructed into a standard QUBO form: If there is a combination constraint (such as a fixed investment amount ), the penalty term can be used for constraint modeling: Considering the transaction fee problem of the initial optimization target, the final optimization problem is represented as: This structure is a typical QUBO problem and is suitable for quantum combinatorial optimizers.

[0048] In the embodiment of the application, the preset quantum optimization algorithm refers to the quantum approximate optimization algorithm (QAOA), which is used to search for an optimal solution in a quantum state space; and the expected energy parameter refers to a maximum expected energy of a quantized optimization target, which is an expected value of a parameterized quantum state under . .

[0049] In the embodiment of the application, the analysis of the expected energy parameter corresponding to the quantum optimization target by using the preset quantum optimization algorithm comprises: Mapping a target variable corresponding to the quantum optimization target to a quantum bit, and mapping the quantum bit to a real number space; constructing a first Hamiltonian and a second Hamiltonian corresponding to the candidate resource combination according to the quantum optimization objective in the real number space; alternately evolving the first Hamiltonian and the second Hamiltonian to obtain a parameterized quantum state; determining an expected energy parameter corresponding to the quantum optimization objective according to a variational parameter in the parameterized quantum state.

[0050] In detail, the variables of the objective function are mapped to quantum bits, and the specific method is to use Ising Model to construct the mapping , so that and satisfy the normalization condition . Since , the corresponding quantum mapping is easy to implement, that is, mapping to , 0 mapping to , and 1 mapping to ; when mapping quantum bits (Hilbert space) to real number space, quantum operators (quantum gates) are usually used to calculate their expected values, and the commonly used method is to introduce Pauli operators, that is, , in order to maintain the corresponding relationship with the quantum mapping, let , so as to realize one-to-one mapping of 0-1 bits and quantum gates (operators), that is, .

[0051] Specifically, the first Hamiltonian (cost Hamiltonian ) is the quantum expression of the quantum optimization objective, which is converted from the standard quadratic form, the commission and the penalty term; the second Hamiltonian (mixed Hamiltonian ) is used for superposition evolution of the quantum state, and the objective function is used. After adding the commission and the penalty term, the cost Hamiltonian using the above mapping method is: the QUBO structure is embedded in the QAOA quantum algorithm for combination search optimization. The cost Hamiltonian is constructed as follows: , wherein , , and after arrangement, the cost Hamiltonian expressed by Ising Model can be obtained: , wherein is a Pauli operator, representing quantum measurement of binary variables.

[0052] Further, by alternately applying corresponding evolution operator and corresponding evolution operator (γ, β are variational parameters), to obtain a parameterized quantum state, that is, QAOA obtains a parameterized quantum state by alternately evolving a cost Hamiltonian and a mixing Hamiltonian: , through the quantum state of such a parameterized circuit, the optimization goal is to maximize the expected energy: , by adjusting , the expected energy parameter is minimized (corresponding to the minimum value of the quantum optimization goal).

[0053] Further, by the parallel search capability of QAOA, the optimization efficiency of high-dimensional combination space is improved, and the problem that traditional methods are difficult to balance generalization and actual constraints is solved.

[0054] S5, analyzing the target quantum state of the expected energy parameter, optimizing the candidate resource combination according to the target quantum state to obtain a target resource combination.

[0055] In the embodiment of the application, the target quantum state refers to a parameterized quantum state that optimizes the expected energy parameter, that is, after optimization , the corresponding .

[0056] In the embodiment of the application, the analysis of the target quantum state of the expected energy parameter comprises: identifying the initial quantum state corresponding to the expected energy parameter; calculating the target resource analysis value of the candidate resource combination corresponding to the initial quantum state; when the target resource analysis value is less than the preset resource analysis threshold, adjusting the expected energy parameter, and returning to the step of identifying the initial quantum state corresponding to the expected energy parameter; when the target resource analysis value is equal to or greater than the preset resource analysis threshold, the initial quantum state corresponding to the expected energy parameter is determined as the target quantum state.

[0057] In detail, the initial quantum state is a set of corresponding , the corresponding expected energy parameter is the minimum value of the current iteration, and the target resource analysis value of the candidate resource combination corresponding to the initial quantum state is measured, when the target resource analysis value is less than the preset resource analysis threshold, the expected energy parameter is adjusted, and the step of identifying the initial quantum state corresponding to the expected energy parameter is returned: the preset resource analysis threshold is a standard for judging whether the combination is optimal, such as 0.06, if the target resource analysis value is 0.05 (less than 0.06), the expected energy parameter is adjusted by a classical optimization algorithm (such as COBYLA), and the initial quantum state is regenerated and the target resource analysis value is calculated.

[0058] ​Specifically, when the target resource analysis value is equal to or greater than the preset resource analysis threshold, the initial quantum state corresponding to the expected energy parameter is determined as the target quantum state, i.e., the initial quantum state is optimized by the classical optimization algorithm COBYLA , and the optimal quantum state of the output result obtained by the final measurement corresponding to the optimal resource combination under the maximum Karnaugh ratio .

[0059] Exemplarily, in a financial scenario, the initial quantum state corresponding to the expected energy parameter -0.5 is identified, and the measurement result is Z=[-1, 1, -1, 1, -1], which is converted into a candidate stock portfolio z=[1, 0, 1, 0, 1] (3 stocks are included). The target resource analysis value (Karnaugh ratio) is calculated to be 0.7, which is greater than the preset threshold 0.6, so the quantum state is determined as the target quantum state, and the corresponding z=[1, 0, 1, 0, 1] is the optimal stock portfolio.

[0060] In addition, in a medical scenario, the initial quantum state corresponding to the expected energy parameter -3.2 is identified, and the measurement result is Z=[-1, 1, 1, -1, 1], which is converted into a candidate drug portfolio z=[1, 0, 0, 1, 1] (2 drugs are used). The target resource analysis value (treatment benefit risk ratio) is calculated to be 4.5, which is greater than the preset threshold 4, so the quantum state is determined as the target quantum state, and the corresponding z=[1, 0, 0, 1, 1] is the optimal drug portfolio.

[0061] As can be seen, in the above scheme, an optimization framework is constructed with the target analysis value prediction function as the core. In view of the characteristics of the target analysis value being nonlinear, non-convex, and non-differentiable, a quantum neural network is used for modeling. Through quadratic regression method, the complex non-convex non-linear prediction target function is converted into a standard format, so that the target analysis value in the form of a ratio can be embedded into the structure of the quantum approximation optimization algorithm, bridging the gap between high-dimensional non-linear modeling and quantum solution, and improving the realizability and operation efficiency of the overall system. Combined with the approximate optimization algorithm, quantum parallel search and quantum entanglement of the combination space are realized, and non-linear optimization is realized. Compared with traditional heuristic algorithms, the solution can be accelerated in a quasi-exponential manner in theory, especially suitable for optimal solution mining in high-dimensional discrete space, which reflects the potential of quantum computing in financial intelligent decision-making, thereby solving the technical problem of low accuracy in resource portfolio optimization.

[0062] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0063] In one embodiment, a quantum network-based resource combination optimization device is provided, which corresponds one-to-one with the quantum network-based resource combination optimization method described in the above embodiments. For example... Figure 5 As shown, the quantum network-based resource combination optimization device includes a target analysis value analysis module 101, a quantum neural network model construction module 102, a quantum function form optimization module 103, a desired energy parameter analysis module 104, and a candidate resource combination optimization module 105. Detailed descriptions of each functional module are as follows: The target analysis value analysis module 101 is used to obtain candidate resource combinations from preset historical transaction data and analyze the target analysis value corresponding to each candidate resource combination. The quantum neural network model construction module 102 is used to construct supervised learning sample pairs based on the candidate resource combination and the target analysis value, and to construct a quantum neural network model of the candidate resource combination using the supervised learning sample pairs; The quantum function form optimization module 103 is used to perform form optimization on the quantum functions of the quantum neural network model to obtain the standard quadratic form; The expected energy parameter analysis module 104 is used to generate the quantum optimization objective of the candidate resource combination according to the standard quadratic form, and to analyze the expected energy parameter corresponding to the quantum optimization objective using a preset quantum optimization algorithm. The candidate resource combination optimization module 105 is used to analyze the target quantum state of the desired energy parameter, optimize the candidate resource combination based on the target quantum state, and obtain the target resource combination.

[0064] In one embodiment, the target analysis value analysis module 101, when performing target analysis value analysis for each candidate resource combination, is used to: Obtain the historical resource data range corresponding to each candidate resource combination; Within a preset time period, extract the expected value and maximum amplitude value corresponding to each candidate resource combination based on the historical resource data interval; Calculate the target analysis value corresponding to each candidate resource combination based on the expected value and the maximum amplitude value.

[0065] In one embodiment, the quantum neural network model building module 102, when executing the construction of a quantum neural network model for the candidate resource combination using the supervised learning sample pairs, is configured to: The candidate resource combinations in the supervised learning samples are mapped to quantum states using a preset angle encoding. The quantum state is nonlinearly mapped to the target analysis value in the supervised learning sample to obtain the initial quantum network model; calculate a target expected value of a pre-acquired quantum state to be analyzed by using the initial quantum network model, and analyze a real expected value corresponding to the quantum state to be analyzed; calculate a loss value between the target expected value and the real expected value, and optimize a circuit parameter in the initial quantum network model according to the loss value; when the loss value is less than a preset loss threshold, use the optimized initial quantum network model as the quantum neural network model of the candidate resource combination.

[0066] In an embodiment, the quantum function form optimization module 103, when performing formal optimization on a quantum function of the quantum neural network model to obtain a standard quadratic form, is configured to: fit a parameter value in the quantum function based on a preset symmetric term quadratic coefficient matrix and a linear term coefficient vector as a benchmark to obtain a fitting result; analyze a fitting error corresponding to the fitting result according to a preset regression target; correct the fitting result according to the fitting error to obtain a standard quadratic form.

[0067] In an embodiment, the expected energy parameter analysis module 104, when generating a quantum optimization target of the candidate resource combination according to the standard quadratic form, is configured to: obtain a combination constraint parameter of the candidate resource combination, and convert the combination constraint parameter into a penalty term parameter of the candidate resource combination; obtain an optimization cost control parameter of an initial optimization target corresponding to the candidate resource combination; combine the penalty term parameter, the optimization cost control parameter, and the standard quadratic form to obtain a quantum optimization target of the candidate resource combination.

[0068] In an embodiment, the expected energy parameter analysis module 104, when analyzing an expected energy parameter corresponding to the quantum optimization target by using a preset quantum optimization algorithm, is configured to: map a target variable corresponding to the quantum optimization target to a quantum bit, and map the quantum bit to a real number space; construct a first Hamiltonian and a second Hamiltonian corresponding to the candidate resource combination in the real number space according to the quantum optimization target; evolve the first Hamiltonian and the second Hamiltonian alternately to obtain a parameterized quantum state; determine the expected energy parameter corresponding to the quantum optimization target according to a variational parameter in the parameterized quantum state.

[0069] In an embodiment, the candidate resource combination optimization module 105, when performing analysis on the target quantum state of the expected energy parameter, is configured to: identify an initial quantum state corresponding to the expected energy parameter; calculate a target resource analysis value of a candidate resource combination corresponding to the initial quantum state; when the target resource analysis value is less than a preset resource analysis threshold, adjust the expected energy parameter and return to the step of identifying the initial quantum state corresponding to the expected energy parameter; when the target resource analysis value is equal to or greater than the preset resource analysis threshold, determine the initial quantum state corresponding to the expected energy parameter as the target quantum state.

[0070] The present application provides a quantum network-based resource combination optimization device, which constructs an optimization framework with a target analysis value prediction function as the core. In view of the characteristics of the target analysis value being nonlinear, non-convex, and non-differentiable, a quantum neural network is used for modeling. A complex non-convex nonlinear prediction target function is converted into a standard format through a quadratic regression method, so that the target analysis value in the form of a ratio can be embedded in the structure of a quantum approximation optimization algorithm, bridging the gap between high-dimensional nonlinear modeling and quantum solving, and improving the realizability and operation efficiency of the overall system. Combined with an approximate optimization algorithm, quantum parallel search and quantum entanglement of the combination space are realized, and nonlinear optimization is realized. Compared with traditional heuristic algorithms, the solution can be accelerated by quasi-exponential in theory, especially suitable for optimal solution mining in high-dimensional discrete space, and reflects the potential of quantum computing in financial intelligent decision-making, thereby solving the technical problem of low accuracy in resource combination optimization.

[0071] The specific limitations of the quantum network-based resource combination optimization device can be referred to the limitations of the quantum network-based resource combination optimization method in the foregoing, which will not be repeated here. Each module in the above quantum network-based resource combination optimization device can be realized by software, hardware, and combinations thereof, in whole or in part. The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module.

[0072] In one embodiment, a computer device is provided, which can be a server, and its internal structure diagram can be as shown in Figure 6As shown in the figure. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes non-volatile and / or volatile storage medium, internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with the external client through the network connection. The computer program is executed by the processor to realize the functions or steps of the server side of the quantum network-based resource combination optimization method.

[0073] In one embodiment, a computer device is provided, which can be a client, and its internal structure diagram can be as shown in the figure. Figure 7 As shown in the figure. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes non-volatile storage medium, internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with the external server through the network connection. The computer program is executed by the processor to realize the functions or steps of the client side of the quantum network-based resource combination optimization method.

[0074] In one embodiment, a computer device is provided, including a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to realize the following steps: Obtain a candidate resource combination in preset historical transaction data, analyze a target analysis value corresponding to each candidate resource combination; According to the candidate resource combination and the target analysis value, a supervised learning sample pair is constructed, and a quantum neural network model of the candidate resource combination is constructed by using the supervised learning sample pair; Formally optimize the quantum function of the quantum neural network model to obtain a standard quadratic form; According to the standard quadratic form, a quantum optimization target of the candidate resource combination is generated, and a preset quantum optimization algorithm is used to analyze an expected energy parameter corresponding to the quantum optimization target; Analyze the target quantum state of the expected energy parameter, and optimize the candidate resource combination according to the target quantum state to obtain a target resource combination.

[0075] In one embodiment, a computer readable storage medium is provided, having stored thereon a computer program which, when executed by a processor, implements the following steps: obtaining a candidate resource combination in preset historical transaction data, analyzing a target analysis value corresponding to each candidate resource combination; constructing a supervised learning sample pair according to the candidate resource combination and the target analysis value, and constructing a quantum neural network model of the candidate resource combination by using the supervised learning sample pair; performing formal optimization on a quantum function of the quantum neural network model to obtain a standard quadratic form; generating a quantum optimization target of the candidate resource combination according to the standard quadratic form, and analyzing an expected energy parameter corresponding to the quantum optimization target by using a preset quantum optimization algorithm; analyzing a target quantum state of the expected energy parameter, and optimizing the candidate resource combination according to the target quantum state to obtain a target resource combination.

[0076] It should be noted that the functions or steps that the computer readable storage medium or the computer device can implement correspond to the related descriptions of the server side and the client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0077] Those skilled in the art can understand that all or part of the processes in the foregoing embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the foregoing embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM) and the like.

[0078] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the above-mentioned division of each functional unit and module is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0079] It should be noted that if non-company software tools or components appear in the embodiments of the present application, they are only used for example introduction and do not represent actual use.

[0080] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A resource combination optimization method based on quantum networks, characterized in that, include: Obtain candidate resource combinations from preset historical transaction data and analyze the target analysis value corresponding to each candidate resource combination; Based on the candidate resource combination and the target analysis value, a supervised learning sample pair is constructed, and the supervised learning sample pair is used to construct a quantum neural network model of the candidate resource combination; The quantum functions of the quantum neural network model are formally optimized to obtain the standard quadratic form; The quantum optimization objective of the candidate resource combination is generated according to the standard quadratic form, and the expected energy parameter corresponding to the quantum optimization objective is analyzed using a preset quantum optimization algorithm; Analyze the target quantum state of the desired energy parameter, and optimize the candidate resource combination based on the target quantum state to obtain the target resource combination.

2. The resource combination optimization method based on quantum networks as described in claim 1, characterized in that, The analysis of the target analysis value corresponding to each candidate resource combination includes: Obtain the historical resource data range corresponding to each candidate resource combination; Within a preset time period, extract the expected value and maximum amplitude value corresponding to each candidate resource combination based on the historical resource data interval; Calculate the target analysis value corresponding to each candidate resource combination based on the expected value and the maximum amplitude value.

3. The resource combination optimization method based on quantum networks as described in claim 1, characterized in that, The construction of the quantum neural network model for the candidate resource combination using the supervised learning sample pairs includes: The candidate resource combinations in the supervised learning samples are mapped to quantum states using a preset angle encoding. The quantum state is nonlinearly mapped to the target analysis value in the supervised learning sample to obtain the initial quantum network model; The target expected value of the quantum state to be analyzed is calculated using the initial quantum network model, and the actual expected value corresponding to the quantum state to be analyzed is analyzed. Calculate the loss value between the target expected value and the actual expected value, and optimize the circuit parameters in the initial quantum network model based on the loss value; When the loss value is less than a preset loss threshold, the optimized initial quantum network model is used as the quantum neural network model for the candidate resource combination.

4. The resource combination optimization method based on quantum networks as described in claim 1, characterized in that, The formal optimization of the quantum function of the quantum neural network model to obtain the standard quadratic form includes: Using a preset symmetric term quadratic coefficient matrix and a linear term coefficient vector as a reference, the parameter values ​​in the quantum function are fitted to obtain the fitting result; Analyze the fitting error corresponding to the fitting result based on the preset regression target; The fitting result is corrected based on the fitting error to obtain the standard quadratic form.

5. The resource combination optimization method based on quantum networks as described in claim 1, characterized in that, The quantum optimization objective for generating the candidate resource combination according to the standard quadratic form includes: Obtain the combination constraint parameters of the candidate resource combination, and convert the combination constraint parameters into penalty term parameters of the candidate resource combination; Obtain the optimization cost control parameters for the initial optimization objective corresponding to the candidate resource combination; The quantum optimization objective of the candidate resource combination is obtained by combining the penalty term parameter, the optimization cost control parameter, and the standard quadratic form.

6. The resource combination optimization method based on quantum networks as described in claim 1, characterized in that, The step of analyzing the expected energy parameter corresponding to the quantum optimization objective using a preset quantum optimization algorithm includes: The target variable corresponding to the quantum optimization objective is mapped to a qubit, and the qubit is mapped to the real number space; In the real number space, construct the first Hamiltonian and the second Hamiltonian corresponding to the candidate resource combination according to the quantum optimization objective; The first Hamiltonian and the second Hamiltonian are alternately evolved to obtain the parameterized quantum state; The desired energy parameter corresponding to the quantum optimization objective is determined based on the variational parameters in the parameterized quantum state.

7. The resource combination optimization method based on quantum networks as described in claim 1, characterized in that, The target quantum state for analyzing the desired energy parameter includes: Identify the initial quantum state corresponding to the desired energy parameter; Calculate the target resource analysis value of the candidate resource combination corresponding to the initial quantum state; When the target resource analysis value is less than the preset resource analysis threshold, the expected energy parameter is adjusted, and the process returns to the step of identifying the initial quantum state corresponding to the expected energy parameter. When the target resource analysis value is equal to or greater than the preset resource analysis threshold, the initial quantum state corresponding to the expected energy parameter is determined as the target quantum state.

8. A resource combination optimization device based on quantum networks, characterized in that, include: The target analysis value analysis module is used to obtain candidate resource combinations from preset historical transaction data and analyze the target analysis value corresponding to each candidate resource combination. A quantum neural network model construction module is used to construct supervised learning sample pairs based on the candidate resource combination and the target analysis value, and to construct a quantum neural network model of the candidate resource combination using the supervised learning sample pairs; The quantum function form optimization module is used to optimize the form of the quantum functions of the quantum neural network model to obtain the standard quadratic form; The expected energy parameter analysis module is used to generate the quantum optimization objective of the candidate resource combination according to the standard quadratic form, and to analyze the expected energy parameter corresponding to the quantum optimization objective using a preset quantum optimization algorithm; The candidate resource combination optimization module is used to analyze the target quantum state of the desired energy parameter, optimize the candidate resource combination based on the target quantum state, and obtain the target resource combination.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the resource combination optimization method based on quantum networks as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the resource combination optimization method based on quantum networks as described in any one of claims 1 to 7.

Images