A performance index real-time evaluation system, method and program product

Abstract

The application discloses a kind of performance index real-time evaluation system, method and program product, belong to artificial intelligence technical field.System, task distributor is configured to N index real-time evaluation agent performance index real-time evaluation task is assigned to task distributor, judge is configured to judge the evaluation result of index real-time evaluation agent and judge, and correct is sent to the performance index real-time evaluation of comprehensive evaluator correct evaluation result carries out performance index real-time evaluation;Comprehensive evaluator correct evaluation result carries out performance index real-time evaluation;Advantage value calculator calculates advantage value according to the evaluation result reported by N index real-time evaluation agent, index real-time evaluation function and reward;The n th index real-time evaluation agent updates the parameter of itself according to advantage value, and carries out performance index real-time evaluation by test data set according to assigned task.The application is fast, and real-time is high.

Description

Technical Field

[0001] This invention relates to a real-time performance evaluation system, method, and program product, belonging to the field of artificial intelligence technology. Background Technology

[0002] Chinese invention patent application CN118114873A discloses an intelligent performance indicator evaluation system. The system includes: an indicator system management module for constructing an indicator evaluation system for the weapons and equipment to be evaluated; an evaluation scheme management module for filtering corresponding marker-level characteristics and the hierarchical relationships and evaluation algorithms between each marker-level characteristic, and associating them to obtain a usable evaluation scheme; a data set management module for filtering characteristic values ​​from the set of hierarchical characteristic values ​​based on the usable evaluation scheme to obtain a usable data set, and associating it to obtain a usable evaluation task; an evaluation task management module for constructing evaluation standards based on one or more weapons and equipment to be evaluated, and associating the evaluation standards with multiple usable evaluation tasks; and an evaluation result generation module for calculating the final evaluation results corresponding to one or more weapons and equipment to be evaluated and displaying them visually.

[0003] While this invention improves the efficiency and accuracy of evaluation, its real-time performance is poor. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a real-time performance evaluation system, method, and computer program product with high real-time performance.

[0005] To achieve the aforementioned objective, this invention provides a real-time performance evaluation system, comprising a central processing unit (CPU) and N real-time performance evaluation agents. The CPU includes a task allocator, a decision maker, a comprehensive evaluator, and an advantage value calculator. The task allocator is configured to assign real-time performance evaluation tasks to the N agents. The decision maker is configured to judge the evaluation results of the agents, transmitting correct results to the comprehensive evaluator for real-time performance evaluation and discarding incorrect results. The comprehensive evaluator is configured to perform real-time performance evaluation based on the correct results provided by the N agents. The advantage value calculator calculates the advantage value based on the evaluation results, evaluation functions, and rewards reported by the agents. The nth agent updates its parameters based on the advantage value, performs real-time performance evaluation using a test dataset according to the assigned task, generates an evaluation result for the assigned task, and reports the evaluation result, evaluation function, and reward to the CPU. n = 1, 2, ..., N.

[0006] To achieve the aforementioned objective, this invention provides a method for real-time evaluation of performance indicators, comprising the following steps:

[0007] Step 1: The task allocator of the central processing unit assigns real-time evaluation tasks to the intelligent agent based on N metrics.

[0008] Step 2: The decision-maker of the central processing unit makes a decision on the evaluation results of the intelligent agent in real time evaluation of the indicators. The correct evaluation results are sent to the comprehensive evaluator for real time evaluation of performance indicators, and the incorrect evaluation results are discarded.

[0009] Step 3: The central processing unit's comprehensive evaluator evaluates the correct evaluation results provided by the intelligent agent in real time based on N indicators.

[0010] Step 4: The advantage value calculator evaluates the agent's reported evaluation results, evaluation function, and reward in real time based on N indicators to calculate the advantage value;

[0011] Step 5: The agent evaluates its parameters in real time based on the advantage value for the nth metric, and performs real-time performance evaluation on the test dataset according to the assigned task, generating an evaluation result for the assigned task, and reporting the evaluation result, evaluation function and reward to the central processing unit, n=1,2,…,N.

[0012] To achieve the aforementioned objective, the present invention also provides a computer program product comprising computer program code capable of being invoked by a processor to execute the methods described above.

[0013] Compared with existing technologies, the real-time performance evaluation method, system, and computer program product provided by this invention have the following beneficial effects:

[0014] This invention evaluates the performance of an intelligent agent in real time using trained metrics, which is fast and highly real-time. Attached Figure Description

[0015] Figure 1 This is a block diagram of the performance index real-time evaluation system provided in the first embodiment of the present invention.

[0016] Figure 2 This is a block diagram of the performance index real-time evaluation system provided in the third embodiment of the present invention. Detailed Implementation

[0017] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0018] First Embodiment

[0019] Figure 1 This is a block diagram of the performance index real-time evaluation system provided in the first embodiment of the present invention, as shown below. Figure 1 As shown, the real-time performance evaluation system provided by this invention includes a central processing unit (CPU) and N real-time performance evaluation agents. The CPU includes a task allocator, a decision maker, a comprehensive evaluator, and an advantage value calculator. The task allocator is configured to assign real-time performance evaluation tasks to the N agents. The decision maker is configured to judge the evaluation results of the agents, sending correct results to the comprehensive evaluator for real-time performance evaluation and discarding incorrect results. The comprehensive evaluator is configured to perform real-time performance evaluation based on the correct evaluation results provided by the N agents. The advantage value calculator calculates the advantage value based on the evaluation results, evaluation functions, and rewards reported by the N agents. The nth agent updates its parameters based on the advantage value, performs real-time performance evaluation using a test dataset according to the assigned task, generates an evaluation result for the assigned task, and reports the evaluation result, evaluation function, and reward to the CPU. n = 1, 2, ..., N.

[0020] The real-time performance evaluation system provided in the first embodiment of the present invention accelerates the calculation speed by using N indicators to evaluate the intelligent agent in real time and calculate multiple indicator values ​​in parallel; and improves the accuracy of the evaluation by setting the decision maker to only evaluate the correct evaluation results provided by the N indicators in real time.

[0021] In the first embodiment, the nth indicator real-time evaluation agent includes an indicator real-time evaluation function. and evaluation function , These are the parameters of the real-time evaluation function for the nth indicator of the intelligent agent. To evaluate the parameters of the evaluation function, the training of the agent for real-time evaluation of the nth metric includes the following process:

[0022] The nth metric is used to evaluate the agent in real time based on test data. Obtain evaluation results and rewards and will Evaluation function and reward The advantage value calculator reported to the central processing unit, among which, ;

[0023] The advantage value calculator calculates the advantage value according to Formula 1. and the advantage value Send to the nth metric for real-time evaluation of the agent:

[0024] Formula 1:

[0025]

[0026] In the formula, This is the discount factor;

[0027] The nth metric is used to evaluate the agent in real time. Update the parameters of the real-time evaluation function for the indicator. .

[0028] In the first embodiment, the agent evaluates the nth indicator in real time using the advantage value according to Formula 2. Update the parameters of the real-time evaluation function for the indicator. for:

[0029] Formula 2:

[0030] ,

[0031] In the formula, Yes gradient function, It is the learning coefficient.

[0032] In the first embodiment, the nth metric evaluates the agent's advantage value in real time. Update the parameters of the evaluation function .

[0033] In the first embodiment, the agent evaluates the nth indicator in real time using the advantage value according to Formula 3. Update the parameters of the evaluation function for:

[0034] Formula 3

[0035] .

[0036] The first embodiment of the present invention obtains the advantage value of the structure above, and evaluates the agent in real time by updating the index through the advantage value, thereby accelerating the convergence speed and training speed.

[0037] In the first embodiment, N real-time evaluation agents provide N index values ​​for evaluating the effectiveness of the equipment, and all N index values ​​are transmitted to the comprehensive evaluator to evaluate the effectiveness of the equipment.

[0038] The comprehensive evaluator evaluates the agent's performance in real time based on N indicators, providing N indicator values ​​for real-time performance evaluation. Specifically, the process includes the following steps:

[0039] S1-1: Based on N indicators, evaluate the correct evaluation results provided by the agent in real time, generate M evaluation schemes, and obtain the evaluation scheme matrix. :

[0040] S1-2: Evaluation scheme matrix The negative indicators are positiveized to obtain an extremely large data matrix. ;

[0041] S1-3: Select the largest number in each column of the extremely large data matrix to form the optimal solution vector. ;

[0042] S1-4: Select the smallest number in each column of the extremely large data matrix to form the worst-case solution vector. ;

[0043] S1-5: For extremely large data matrices Normalizing each element in the matrix yields a normalized extremely large data matrix. In the formula, ;

[0044] S1-6: For the m-th solution, calculate its distance from 1: ; Calculate its relationship with Distance to the optimal solution: ;

[0045] S1-7: Calculate the score of the m-th scheme: ;

[0046] S1-8: Based on the score Sort and score. The highest-performing option will be used as the performance evaluation scheme.

[0047] Optionally, steps S1-7 and S1-8 can be replaced by steps S1-9 and S1-10 respectively:

[0048] S1-9: Calculate the score of the m-th scheme: ;

[0049] S1-10: Based on the score Sort and score. The smallest possible solution is used as the performance evaluation solution.

[0050] The comprehensive evaluator provided in the first embodiment of the present invention can obtain the optimal technical solution for performance evaluation through the above technical solution.

[0051] The real-time performance evaluation system provided in the first embodiment also includes an input unit, through which users input instructions, test data, etc.

[0052] The performance indicator real-time evaluation system provided in the first embodiment also includes an output unit, which includes a display, etc. The display is used to display test data, real-time evaluation function of indicators, evaluation function, evaluation results, etc.

[0053] In the first embodiment of the present invention, a real-time performance evaluation agent is used to evaluate performance indicators in real time. This agent is capable of autonomous learning and has the ability to adapt to new test data and new indicator real-time evaluation tasks.

[0054] The first embodiment of the present invention does not require the use of an experience pool to store historical samples, thus saving storage space and improving data sampling efficiency, thereby increasing training speed.

[0055] Second Embodiment

[0056] The second embodiment of the present invention only describes the contents that are different from those of the first embodiment; the contents that are the same will not be described again.

[0057] The second embodiment of the present invention provides a method for real-time evaluation of performance indicators, which includes the following steps:

[0058] Step 1: The task allocator of the central processing unit assigns real-time evaluation tasks to the intelligent agent based on N metrics.

[0059] Step 2: The decision-maker of the central processing unit makes a decision on the evaluation results of the intelligent agent in real time evaluation of the indicators. The correct evaluation results are sent to the comprehensive evaluator for real time evaluation of performance indicators, and the incorrect evaluation results are discarded.

[0060] Step 3: The central processing unit's comprehensive evaluator evaluates the correct evaluation results provided by the intelligent agent in real time based on N indicators.

[0061] Step 4: The advantage value calculator evaluates the agent's reported evaluation results, evaluation function, and reward in real time based on N indicators to calculate the advantage value;

[0062] Step 5: The agent evaluates its parameters in real time based on the advantage value for the nth metric, and performs real-time performance evaluation on the test dataset according to the assigned task, generating an evaluation result for the assigned task, and reporting the evaluation result, evaluation function and reward to the central processing unit, n=1,2,…,N.

[0063] In the second embodiment of the present invention, a real-time performance evaluation agent is used to evaluate performance indicators in real time. This agent is capable of learning autonomously and has the ability to adapt to new data resources and new performance indicator real-time evaluation tasks.

[0064] The second embodiment of the present invention eliminates the need to store historical samples in an experience pool, saving storage space and improving data sampling efficiency, thereby increasing training speed. Simultaneously, by collecting samples from multiple different training environments, the sample distribution is more uniform, which is more conducive to neural network training.

[0065] Third Embodiment

[0066] Figure 2 This is a block diagram of the performance index real-time evaluation system provided in the third embodiment of the present invention, as shown below. Figure 2 As shown, the real-time performance evaluation system provided by this invention includes a central processing unit and N real-time performance evaluation agents. The central processing unit includes a task allocator, a decision-maker, and a comprehensive evaluator. The task allocator is configured to assign real-time performance evaluation tasks to the N real-time performance evaluation agents. The decision-maker is configured to judge the evaluation results of the real-time performance evaluation agents, transmit the correct evaluation results to the comprehensive evaluator for real-time performance evaluation, and discard the incorrect evaluation results. The comprehensive evaluator is configured to perform real-time performance evaluation based on the correct evaluation results provided by the N real-time performance evaluation agents.

[0067] The real-time evaluation intelligence for the nth indicator includes an advantage value calculator. The advantage value calculator calculates the advantage value based on the evaluation results, evaluation function, and reward reported by the real-time evaluation intelligence for the nth indicator. The real-time evaluation intelligence for the nth indicator updates its own parameters based on the advantage value and performs real-time evaluation of the indicator through the test dataset according to the assigned task, generating an evaluation result for the assigned task, n=1,2,…,N.

[0068] The real-time performance evaluation system provided in the third embodiment of the present invention accelerates the calculation speed by using N indicators to evaluate the intelligent agent in real time and calculate multiple indicator values ​​in parallel; and improves the accuracy of the evaluation by setting the decision maker to only evaluate the correct evaluation results provided by the N indicators in real time.

[0069] In the third embodiment, the nth indicator real-time evaluation agent includes an indicator real-time evaluation function. and evaluation function , These are the parameters of the real-time evaluation function for the nth indicator of the intelligent agent. , The training of the agent for real-time evaluation of the nth metric, denoted as the parameters of the value function, includes the following process:

[0070] The nth metric is used to evaluate the agent in real time based on test data. Obtain evaluation results and rewards The advantage value calculator uses formula 1 to... Evaluation function and reward Calculate the advantage value ;

[0071] The nth metric is used to evaluate the agent in real time. Update the parameters of the real-time evaluation function for the indicator. .

[0072] In the third embodiment, the agent evaluates the nth indicator in real time using the advantage value according to Formula 2. Update the parameters of the real-time evaluation function for the indicator. .

[0073] In the third embodiment, the nth metric evaluates the agent's advantage value in real time. Update the parameters of the evaluation function .

[0074] In the third embodiment, the agent evaluates the nth indicator in real time using the advantage value according to Formula 3. Update the parameters of the evaluation function .

[0075] The first embodiment of the present invention obtains the advantage value of the structure above, and evaluates the agent in real time by updating the index through the advantage value, thereby accelerating the convergence speed and training speed.

[0076] In the first embodiment, N real-time evaluation agents provide N index values ​​for evaluating the effectiveness of the equipment, and all N index values ​​are transmitted to the comprehensive evaluator to evaluate the effectiveness of the equipment.

[0077] The comprehensive evaluator provided in the third embodiment of the present invention can obtain the optimal technical solution for performance evaluation through the above technical solution.

[0078] The real-time performance evaluation system provided in the third embodiment also includes an input unit, through which users input instructions, test data, etc.

[0079] The real-time performance evaluation system provided in the third embodiment also includes an output unit, which includes a display, etc. The display is used to display test data, real-time performance evaluation function, evaluation function, evaluation results, etc.

[0080] In the third embodiment of the present invention, a real-time performance evaluation agent is used to evaluate performance indicators in real time. This agent is capable of autonomous learning and has the ability to adapt to new test data and new indicator real-time evaluation tasks.

[0081] The third embodiment of the present invention does not require the use of an experience pool to store historical samples, thus saving storage space and improving data sampling efficiency, thereby increasing training speed.

[0082] Fourth embodiment

[0083] The fourth embodiment of the present invention only describes the content that is different from the third embodiment; the same content will not be described again.

[0084] The fourth embodiment of the present invention provides a method for real-time evaluation of performance indicators, which includes the following steps:

[0085] S2-1: The task allocator of the central processing unit assigns real-time evaluation tasks to N metrics for the intelligent agent.

[0086] S2-2: The decision-maker of the central processing unit makes a decision on the evaluation results of the intelligent agent in real time evaluation of the indicators. The correct evaluation results are transmitted to the comprehensive evaluator for real time evaluation of performance indicators, and the incorrect evaluation results are discarded.

[0087] S2-3: The central processing unit's comprehensive evaluator evaluates the correct evaluation results provided by the intelligent agent in real time based on N indicators.

[0088] S2-4: The nth indicator real-time evaluation agent evaluates the performance indicators in real time through the test dataset according to the assigned task, generates the evaluation result for the assigned task, and calculates the advantage value according to the evaluation result, evaluation function and reward, and updates its own parameters n=1,2,…,N according to the advantage value.

[0089] In the fourth embodiment of the present invention, a real-time performance evaluation agent is used to evaluate performance indicators in real time. This agent is capable of learning autonomously and has the ability to adapt to new data resources and new performance indicator real-time evaluation tasks.

[0090] The fourth embodiment of the present invention eliminates the need for storing historical samples in an experience pool, saving storage space and improving data sampling efficiency, thereby increasing training speed. Simultaneously, by collecting samples from multiple different training environments, the sample distribution is more uniform, which is more conducive to neural network training.

[0091] Fifth embodiment

[0092] The fifth embodiment of the present invention provides a computer program product, characterized in that it includes computer program code, which can be called by a processor to execute the method described in the second embodiment or the fourth embodiment.

[0093] The beneficial effects of the fifth embodiment of the present invention are the same as those of the second or fourth embodiment, and will not be described again here.

[0094] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. "Several" means one or more, unless otherwise explicitly specified.

[0095] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A real-time performance indicator evaluation system, characterized in that, The system comprises a central processing unit (CPU) and N real-time evaluation agents for various metrics. The CPU includes a task allocator, a decision maker, a comprehensive evaluator, and an advantage value calculator. The task allocator is configured to assign real-time evaluation tasks to the N agents. The decision maker is configured to judge the evaluation results of the agents, sending correct results to the comprehensive evaluator for real-time performance evaluation and discarding incorrect results. The comprehensive evaluator is configured to perform real-time performance evaluation based on the correct results provided by the N agents. The advantage value calculator calculates the advantage value based on the evaluation results, evaluation functions, and rewards reported by the agents. The nth agent updates its parameters based on the advantage value, performs real-time evaluation of metrics using a test dataset according to the assigned task, generates an evaluation result for the assigned task, and reports the evaluation result, evaluation function, and reward to the CPU. (n=1,2,…,N). The real-time evaluation agent for the nth indicator includes the real-time evaluation function for the indicator. and evaluation function , These are the parameters of the real-time evaluation function for the nth indicator of the intelligent agent. To evaluate the parameters of the evaluation function, the training of the agent for real-time evaluation of the nth metric includes the following process: The nth metric is used to evaluate the agent in real time based on test data. Obtain evaluation results and rewards and the evaluation results Evaluation function and reward The advantage value calculator reported to the central processing unit, among which, ; The advantage value calculator calculates the advantage value using the following formula. and the advantage value Send to the nth metric for real-time evaluation of the agent: ， In the formula, This is the discount factor; The nth metric is used to evaluate the agent in real time. Update the parameters of the real-time evaluation function for the indicator. : ， In the formula, Yes gradient function, It is the learning coefficient.

2. The real-time performance indicator evaluation system according to claim 1, characterized in that, The nth metric for real-time evaluation of the agent also uses the advantage value. Update the parameters of the evaluation function : 。 3. A method for real-time evaluation of performance indicators, characterized in that, Includes the following steps: Step 1: The task allocator of the central processing unit assigns real-time evaluation tasks to the intelligent agent based on N metrics. Step 2: The decision-maker of the central processing unit makes a decision on the evaluation results of the intelligent agent in real time evaluation of the indicators. The correct evaluation results are sent to the comprehensive evaluator for real time evaluation of performance indicators, and the incorrect evaluation results are discarded. Step 3: The central processing unit's comprehensive evaluator evaluates the correct evaluation results provided by the intelligent agent in real time based on N indicators. Step 4: The advantage value calculator evaluates the agent's reported evaluation results, evaluation function, and reward in real time based on N indicators to calculate the advantage value; Step 5: The agent evaluates its parameters in real time based on the advantage value for the nth metric, and performs real-time performance evaluation on the test dataset according to the assigned task, generating an evaluation result for the assigned task, and reporting the evaluation result, evaluation function and reward to the central processing unit, n=1,2,…,N; The real-time evaluation agent for the nth indicator includes the real-time evaluation function for the indicator. and evaluation function , These are the parameters of the real-time evaluation function for the nth indicator of the intelligent agent. To evaluate the parameters of the evaluation function, the training of the agent for real-time evaluation of the nth metric includes the following process: The nth metric is used to evaluate the agent in real time based on test data. Obtain evaluation results and rewards and the evaluation results Evaluation function and reward The advantage value calculator reported to the central processing unit, among which, ; The advantage value calculator calculates the advantage value using the following formula. and the advantage value The data is transmitted to the nth metric for real-time evaluation of the agent. ， In the formula, This is the discount factor; The nth metric is used to evaluate the agent in real time. Update the parameters of the real-time evaluation function for the indicator. : ， In the formula, Yes gradient function, It is the learning coefficient.

4. The real-time performance indicator evaluation method according to claim 3, characterized in that, The nth metric for real-time evaluation of the agent also uses the advantage value. Update the parameters of the evaluation function : 。 5. A computer program product, characterized in that, It includes computer program code that can be called by a processor to execute the real-time performance evaluation method according to any one of claims 3-4.

Images