A method and apparatus for optimizing a communication structure

Abstract

The application discloses a communication structure optimization method, which comprises the following steps: determining a first intelligent agent group, the first intelligent agent group comprising a plurality of intelligent agents having a cooperative relationship in a first reasoning process of information to be processed; determining a first evaluation index, the first evaluation index being used for indicating the cooperative effect of all intelligent agents in the first intelligent agent group; and determining whether to retain the first intelligent agent group based on the first evaluation index. Through the above method, the accuracy of task processing can be improved.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a method and apparatus for optimizing communication structures. Background Technology

[0002] With the development of large language models, multi-agent systems are widely used in various complex reasoning tasks. By relying on multiple agents with different roles and abilities to cooperate, the accuracy of task processing is improved.

[0003] Currently, multi-agent systems typically pre-build a point-to-point agent communication structure. Then, during model training or task execution, the multi-agent system optimizes the agent communication structure by removing low-value communication edges or agents with low contributions, thereby compressing communication links, reducing information redundancy, and ultimately forming a sparse point-to-point agent communication structure.

[0004] However, the accuracy of task processing using the optimized communication structure derived from this method is low. Summary of the Invention

[0005] This application provides a communication structure optimization method to improve the accuracy of task processing using the optimized communication structure. Furthermore, this application also provides a corresponding communication structure optimization apparatus, computer equipment, computer-readable storage medium, and computer program product.

[0006] In a first aspect, this application provides a communication structure optimization method, the method comprising: determining a first agent group, the first agent group including multiple agents that have a cooperative relationship during the first reasoning process of information to be processed; determining a first evaluation index, the first evaluation index being used to indicate the cooperative effect of all agents in the first agent group; and determining whether to retain the first agent group based on the first evaluation index.

[0007] In one possible implementation, the method further includes: determining a first agent in the first agent group if the first evaluation metric indicates that the first agent group should be retained; wherein the first agent is used to summarize multiple inference results output by all agents in the first agent group to obtain a first sub-inference result, and one agent outputs one inference result.

[0008] In one possible implementation, the method further includes: sending the first sub-inference result to a second agent, the second agent being used to summarize multiple inference results output by all agents in the second agent group to obtain a second sub-inference result, wherein the first agent group and the second agent group have a communication connection, and the second agent group includes multiple agents that have a cooperative relationship in the first inference process of the information to be processed; and using the second agent to infer the first sub-inference result and the second sub-inference result to obtain the first inference result.

[0009] In one possible implementation, the method further includes: determining a final reasoning result for the information to be processed based on the first reasoning result and the second reasoning result, wherein the second reasoning result is used to instruct a third agent to summarize the reasoning results of all agents in a third agent group to obtain a third sub-reasoning result; wherein the third agent group includes multiple agents that have a cooperative relationship in the first reasoning process of the information to be processed, and the third agent group has no communication connection with either the first agent group or the second agent group.

[0010] In one possible implementation, the method further includes: determining a fourth agent group, the fourth agent group comprising multiple agents that have a cooperative relationship during the second reasoning process of the information to be processed; determining a fourth evaluation index based on the final reasoning result, the fourth evaluation index being used to indicate the cooperative effect of all agents in the fourth agent group; and determining whether to retain the fourth agent group based on the fourth evaluation index.

[0011] In one possible implementation, determining the first evaluation index includes: determining the first evaluation index based on a first parameter, a second parameter, a third parameter, and a fourth parameter; wherein the first parameter is used to indicate the number of times the first agent group successfully infers the information to be processed, the second parameter is used to indicate the number of times the first agent group fails to infer the information to be processed, the third parameter is used to indicate the evaluation score of the agent communication edge in the first agent group, the agent communication edge is used to indicate the existence of a communication relationship between two agents, and the fourth parameter is used to indicate the degree of duplication between the agents in the first agent group and the agents in the retained agent group.

[0012] Secondly, this application provides a communication structure optimization device, the device comprising: a first determining module, configured to determine a first intelligent agent group, the first intelligent agent group including multiple intelligent agents that have a cooperative relationship during the first reasoning process of the information to be processed; a second determining module, configured to determine a first evaluation index, the first evaluation index being used to indicate the cooperative effect of all intelligent agents in the first intelligent agent group; and a third determining module, configured to determine whether to retain the first intelligent agent group based on the first evaluation index.

[0013] In one possible implementation, the apparatus further includes: a fourth determining module, configured to determine a first agent in the first agent group when the first evaluation index indicates that the first agent group should be retained; wherein the first agent is configured to summarize multiple inference results output by all agents in the first agent group to obtain a first sub-inference result, with one agent outputting one inference result.

[0014] In one possible implementation, the apparatus further includes: a sending module, configured to send the first sub-inference result to a second agent, the second agent being configured to aggregate multiple inference results output by all agents in the second agent group to obtain a second sub-inference result, the first agent group and the second agent group having a communication connection, the second agent group including multiple agents having a cooperative relationship in the first inference process of the information to be processed; and using the second agent to infer the first sub-inference result and the second sub-inference result to obtain the first inference result.

[0015] In one possible implementation, the apparatus further includes: a fifth determining module, configured to determine the final reasoning result of the information problem to be processed based on the first reasoning result and the second reasoning result, wherein the second reasoning result is used to instruct a third agent to summarize the reasoning results of all agents in the third agent group to obtain a third sub-reasoning result; wherein the third agent group includes multiple agents that have a cooperative relationship in the first reasoning process of the information to be processed, and the third agent group has no communication connection with either the first agent group or the second agent group.

[0016] In one possible implementation, the apparatus further includes: a sixth determining module, configured to determine a fourth agent group, the fourth agent group comprising multiple agents that have a cooperative relationship during the second reasoning process of the information to be processed; determine a fourth evaluation index based on the final reasoning result, the fourth evaluation index being used to indicate the cooperative effect of all agents in the fourth agent group; and determine whether to retain the fourth agent group based on the fourth evaluation index.

[0017] In one possible implementation, the second determining module is configured to determine the first evaluation index based on a first parameter, a second parameter, a third parameter, and a fourth parameter; wherein the first parameter indicates the number of times the first agent group successfully infers the information to be processed, the second parameter indicates the number of times the first agent group fails to infer the information to be processed, the third parameter indicates the evaluation score of the agent communication edge in the first agent group, the agent communication edge indicates the existence of a communication relationship between two agents, and the fourth parameter indicates the degree of duplication between the agents in the first agent group and the agents in the retained agent group.

[0018] Thirdly, this application provides a computer device including a processor and a memory, the memory being used to store instructions, and the processor executing the instructions stored in the memory to cause the computer device to perform any of the aforementioned communication structure optimization methods.

[0019] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the communication structure optimization method as described in any of the above.

[0020] Fifthly, this application provides a computer program product containing instructions that, when run on at least one computing device, cause the at least one computing device to perform any of the aforementioned communication structure optimization methods.

[0021] Based on the implementation methods provided in the above aspects, this application can be further combined to provide more implementation methods.

[0022] As can be seen from the above technical solutions, this application has the following beneficial effects: The communication structure optimization system acquires information to be processed and identifies multiple agents to handle that information. Subsequently, the system utilizes these agents to perform an initial inference on the information and, based on the communication relationships among the agents during this initial inference, identifies multiple agent groups with cooperative relationships. Furthermore, the system can determine evaluation metrics for each agent group and, based on these metrics, decide whether to retain that agent group. In this way, the communication structure optimization system can eliminate inefficient collaborative agent groups and deeply compress communication links. Compared with the communication structure optimization system that optimizes the communication structure based on the evaluation index of a single communication edge, the embodiments of this application upgrade the optimization object from scattered single communication edges to complete collaborative agent groups. This not only can batch clean up all communication connections corresponding to inefficient agent groups and significantly reduce information redundancy in the multi-agent system, but also avoid the destruction of the inherent collaborative associations within the agent groups by edge-by-edge deletion operations. While achieving lightweight and sparse optimization of the communication structure, it fully preserves the interaction logic and working capabilities of high-quality collaborative agent groups, effectively improving the communication transmission efficiency and collaborative inference performance of the multi-agent system. The optimized communication structure can also better adapt to the operational requirements of actual tasks and improve the accuracy of task processing. Attached Figure Description

[0023] Figure 1 A flowchart illustrating a communication structure optimization method provided in an embodiment of this application; Figure 2 A schematic diagram of a hyperedge structure provided in an embodiment of this application; Figure 3 A schematic diagram illustrating inter-agent communication provided in an embodiment of this application; Figure 4 Another flowchart illustrating a communication structure optimization method provided in this application embodiment; Figure 5 This is a schematic diagram of a communication structure optimization device provided in an embodiment of this application. Detailed Implementation

[0024] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate; this is merely a method of distinction used in describing objects with the same attributes in the embodiments of this application.

[0025] To make the above-mentioned objectives, features and advantages of the embodiments of this application more apparent and understandable, the embodiments of this application will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0026] With the development of large language models, multi-agent systems are widely used in various complex reasoning tasks. By relying on multiple agents with different roles and abilities to cooperate, the accuracy of task processing is improved.

[0027] Currently, multi-agent systems typically pre-build a point-to-point agent communication structure. Then, during model training or task execution, the multi-agent system optimizes the agent communication structure by removing low-value communication edges or agents with low contributions, thereby compressing communication links, reducing information redundancy, and ultimately forming a sparse point-to-point agent communication structure.

[0028] However, the accuracy of task processing using the optimized communication structure derived from this method is low.

[0029] Based on this, embodiments of this application provide a communication structure optimization method, which optimizes the communication structure based on the evaluation index of intelligent agent groups to improve the accuracy of task processing.

[0030] In practice, the communication structure optimization system acquires the information to be processed and identifies multiple agents to handle it. Subsequently, the system uses these agents to perform an initial inference on the information and, based on the communication relationships between the agents during this initial inference, identifies multiple agent groups with cooperative relationships. Further, the system determines evaluation metrics for each agent group and decides whether to retain that group based on these metrics. In this way, the communication structure optimization system can eliminate inefficient collaborative agent groups and deeply compress communication links. Compared with the communication structure optimization system that optimizes the communication structure based on the evaluation index of a single communication edge, the embodiments of this application upgrade the optimization object from scattered single communication edges to complete collaborative agent groups. This not only can batch clean up all communication connections corresponding to inefficient agent groups and significantly reduce information redundancy in the multi-agent system, but also avoid the destruction of the inherent collaborative associations within the agent groups by edge-by-edge deletion operations. While achieving lightweight and sparse optimization of the communication structure, it fully preserves the interaction logic and working capabilities of high-quality collaborative agent groups, effectively improving the communication transmission efficiency and collaborative inference performance of the multi-agent system. The optimized communication structure can also better adapt to the operational requirements of actual tasks and improve the accuracy of task processing.

[0031] To facilitate understanding of the embodiments of this application, a communication structure optimization method provided by the embodiments of this application will be described below with reference to the accompanying drawings.

[0032] See Figure 1 , Figure 1 This is a schematic diagram of a communication structure optimization process provided in an embodiment of this application.

[0033] in, Figure 1 The communication structure optimization method shown may specifically include the following steps.

[0034] S101: The communication structure optimization system determines the first intelligent agent group, which includes multiple intelligent agents that have a cooperative relationship during the first reasoning process of the information to be processed.

[0035] In its implementation, the communication structure optimization system first receives user-inputted information to be processed. This information may include the problem to be processed, the business domain, the target number, and the basic model configuration. The business domain refers to the business domain to which the problem belongs, and the target number is the number of agents needed to process the problem. The basic model configuration indicates the type of large model invoked by the agents.

[0036] Then, the communication structure optimization system can determine target agents from the role capability library based on the business domain and target quantity in the information to be processed. The target agents can handle issues related to the business domain in the information to be processed. The target agents can include multiple agents, the number of which is the target quantity. Subsequently, the target agents can invoke a larger model based on the basic model configuration in the information to be processed, in order to perform reasoning on the problem to be processed and obtain the reasoning result.

[0037] It should be noted that the role capability library can include multiple agents, each of which has a clear role and cue words, such as legal strategist, scientific reasoner, medical assessor, and general analyst.

[0038] Furthermore, the communication structure optimization system can input the problem to be processed into the target agent, which then performs an initial reasoning operation to obtain a reasoning result. During this initial reasoning process, the communication structure optimization system can save the information transmission relationships between the target agents. Subsequently, based on these information transmission relationships, the system can determine a first agent group, which may include multiple agents that have cooperative relationships during the initial reasoning process of the problem.

[0039] Similarly, the communication structure optimization system can determine the second and third agent groups based on the information transmission relationship between the target agents. The second agent group can include multiple agents that have a cooperative relationship in the first reasoning of the problem to be processed, and the third agent group can include multiple agents that have a cooperative relationship in the first reasoning of the problem to be processed.

[0040] It should be noted that during this process, the communication structure optimization system can identify two, three, or more groups of intelligent agents, without specifying the number of groups of intelligent agents.

[0041] It is worth noting that the same agent can exist in both the first agent group and the second agent group simultaneously; that is, agents can be duplicated in both groups. For example, the first agent group may include Agent 1, Agent 2, and Agent 3. The second agent group may include Agent 1, Agent 2, and Agent 4. Alternatively, agents may not be duplicated in either the first or second agent group. For example, the first agent group may include Agent 1, Agent 2, and Agent 3. The second agent group may include Agent 4, Agent 5, and Agent 6.

[0042] In one possible implementation, the first, second, and third agent groups can be hyperedges. Hyperedges can be used to characterize topological units that can connect two or more agents, and to model cooperative agent groups composed of multiple agents, as well as information interaction and collaborative reasoning behaviors within the agent groups.

[0043] See Figure 2 , Figure 2 A schematic diagram of a hyperedge structure is shown.

[0044] S102: The communication structure optimization system determines a first evaluation index, which is used to indicate the cooperative effect of all agents in the first agent group.

[0045] After determining the first, second, and third agent groups, the communication structure optimization system can determine a first evaluation index for the first agent group, a second evaluation index for the second agent group, and a third evaluation index for the third agent group. The first evaluation index can be used to indicate the collaborative effect of multiple agents in the first agent group, the second evaluation index can be used to indicate the collaborative effect of multiple agents in the second agent group, and the third evaluation index can be used to indicate the collaborative effect of multiple agents in the third agent group.

[0046] In one example, the first, second, and third evaluation metrics can be evaluation scores. The communication structure optimization system can determine the first, second, and third evaluation metrics based on the first, second, third, and fourth parameters. Specifically, the first parameter can indicate the number of times the first agent group successfully infers the information to be processed; the second parameter can indicate the number of times the first agent group fails to infer the information to be processed; the third parameter can indicate the evaluation score of the agent communication edge in the first agent group; and the fourth parameter indicates the degree of overlap between the first agent group and the retained agents. The agent communication edge can characterize the communication relationship between two agents.

[0047] The specific expression for the evaluation index is as follows: ; in, Characterizing intelligent agent groups; , , , These are the weighting coefficients; For the evaluation index of agent group e; The number of inferences required for an agent group e to obtain a correct inference result from the information to be processed. The number of inferences that fail to yield the correct result for the information to be processed by the agent group e; is the average evaluation score of all communication edges among agents in agent group e; The degree of overlap between agent group e and agents in the retained agent groups.

[0048] The evaluation score of the agent's communication edge can be determined according to the following formula: ; ; in, The score represents the number of communication edges between the agents. The vectorized information to be processed from one of the agents is input into the agent communication edge; The vectorized information to be processed is input into the communication edge between the agents; n is the dimension of the vectorized information to be processed.

[0049] S103: The communication structure optimization system determines whether to retain the first intelligent agent group based on the first evaluation index.

[0050] In a first possible implementation, the communication structure optimization system can determine whether to retain the first intelligent agent group, the second intelligent agent group, and the third intelligent agent group according to a preset ratio.

[0051] Specifically, the communication structure optimization system can pre-set a preset ratio. The system can first sort the first and second agent groups according to their evaluation scores from highest to lowest, obtaining a ranking result. Then, the system can retain agent groups from highest to lowest based on the preset ratio and the ranking result.

[0052] For example, a preset ratio of 2 / 3 means that two agent groups should be retained from the first agent group, the second agent group, and the third agent group. If the evaluation score of the first agent group is higher than that of the second agent group, and the evaluation score of the second agent group is higher than that of the third agent group, then the communication structure optimization system can retain the first and second agent groups, but not the third agent group.

[0053] In a second possible implementation, the communication structure optimization system can determine whether to retain the first agent group, the second agent group, and the third agent group according to a preset score threshold.

[0054] Specifically, the communication structure optimization system can pre-set a preset score threshold. If the first evaluation score is not less than the preset score threshold, the communication structure optimization system can retain the first agent group. If the first evaluation score is less than the preset score threshold, the communication structure optimization system can discontinue the first agent group. Similarly, if the second evaluation score is not less than the preset score threshold, the communication structure optimization system can retain the second agent group. If the second evaluation score is less than the preset score threshold, the communication structure optimization system can discontinue the second agent group. If the third evaluation score is not less than the preset score threshold, the communication structure optimization system can retain the third agent group. If the third evaluation score is less than the preset score threshold, the communication structure optimization system can discontinue the third agent group.

[0055] The communication structure optimization system, through the methods provided in the above embodiments, can divide intelligent agent groups based on the collaborative relationships formed during the inference process, and evaluate the overall collaborative effect of multiple intelligent agents within the intelligent group using evaluation metrics. Based on the overall evaluation, it determines whether to retain the intelligent agent group. In this way, the communication structure optimization system can eliminate inefficient collaborative intelligent agent groups and deeply compress communication links. Compared to communication structure optimization systems that optimize communication structures based on evaluation metrics of single communication edges, this embodiment upgrades the optimization object from scattered single communication edges to complete collaborative intelligent agent groups. This not only allows for the batch cleaning of all communication connections corresponding to inefficient intelligent agent groups, significantly reducing information redundancy within the multi-agent system, but also avoids edge-by-edge deletion operations that could disrupt the inherent collaborative relationships within the intelligent agent groups. While achieving lightweight and sparse optimization of the communication structure, it fully preserves the interaction logic and working capabilities of high-quality collaborative intelligent agent groups, effectively improving the communication transmission efficiency and collaborative inference performance of the multi-agent system. The optimized communication structure can also better adapt to the operational needs of actual tasks and improve the accuracy of task processing.

[0056] In one possible implementation, when the communication structure optimization system simultaneously retains a first group of agents, a second group of agents, and a third group of agents, the communication structure optimization system can determine the first agent in the first group of agents, the second agent in the second group of agents, and the third agent in the third group of agents.

[0057] Specifically, the communication structure optimization system can determine a representative agent from the agent group using the following formula.

[0058]

[0059]

[0060] in, The determined agent represents For the i-th intelligent agent; For intelligent agent groups; The score represents the communication edge between the i-th agent and the j-th agent. The score represents the communication edge between the j-th agent and the i-th agent. These are the weighting coefficients; This represents the predefined probability that the i-th agent is the agent representative.

[0061] Furthermore, the first agent can summarize the reasoning results output by all agents in the first agent group to obtain a first sub-reasoning result. The second agent can summarize the reasoning results output by all agents in the second agent group to obtain a second sub-reasoning result. The third agent can summarize the reasoning results output by all agents in the third agent group to obtain a third sub-reasoning result. Each agent outputs one reasoning result.

[0062] It is understandable that the first intelligent agent, the second intelligent agent, and the third intelligent agent are the representative nodes of the first intelligent agent group, the second intelligent agent group, and the third intelligent agent group, respectively.

[0063] It should be noted that the first intelligent agent group includes the first intelligent agent, the second intelligent agent group includes the second intelligent agent, and the third intelligent agent group includes the third intelligent agent.

[0064] In one possible implementation, groups of agents can also communicate according to a directed acyclic graph. See also Figure 3 , Figure 3 A schematic diagram of inter-agent communication is shown. (For example...) Figure 3 As shown, the first agent group and the second agent group have a communication relationship, with the first agent group being upstream of the second agent group. The third agent group has no communication relationship with either the first or second agent groups. The communication structure optimization system can also send the first sub-inference result of the first agent group to the second agent. The second agent can perform reasoning on the first and second sub-inference results to update the second sub-inference result and obtain the first inference result. It can be understood that the first inference result is the joint inference result of the first and second agent groups. Since the third agent group does not communicate with other agent groups, the communication structure optimization system can determine the third sub-inference result of the third agent group as the second inference result.

[0065] Furthermore, the communication structure optimization system can aggregate the reasoning results of all agent groups to obtain the final reasoning result of the information to be processed.

[0066] In a first possible implementation, the communication structure optimization system can define a final decision agent. This final decision agent can aggregate the inference results of all agents located at the end nodes of the directed acyclic graph to obtain the final processed result of the information to be processed. For example, the communication structure optimization system can aggregate the first inference result and the second inference result to obtain the final processed result of the information to be processed.

[0067] In a second possible implementation, the communication structure optimization system can set up a final decision agent. The final decision agent can further summarize the reasoning results of agent groups whose evaluation index is greater than a preset threshold, based on the summarization of the reasoning results of all agent groups located at the end nodes of the directed acyclic graph, so as to reduce information loss in the process of inter-group information transmission.

[0068] Furthermore, the communication structure optimization system can compare the aggregated final inference result with the standard inference result to verify whether the final inference result is correct. If the final inference result is correct, the communication structure optimization system can increment the number of successful inferences of the agent group that participated in obtaining the final inference result by one; otherwise, the communication structure optimization system can increment the number of failed inferences of the agent group that participated in obtaining the final inference result by one.

[0069] See Figure 4 , Figure 4 Another flowchart of a communication structure optimization method is shown.

[0070] In one possible implementation, after the communication structure optimization system obtains the final reasoning result of the first reasoning of the information to be processed, it can perform a second optimization of the communication structure.

[0071] Specifically, the communication structure optimization system can perform a second reasoning on the information to be processed, and can determine a fourth intelligent agent group based on the communication relationships existing in the second reasoning process. The fourth intelligent agent group includes multiple intelligent agents that have cooperative relationships in the second reasoning process of the information to be processed.

[0072] Then, the communication structure optimization system can determine a fourth evaluation metric based on the final inference result of the first inference. This fourth evaluation metric indicates the collaborative effectiveness of all agents in the fourth agent group. That is, the communication structure optimization system updates the number of successful or failed inferences by the fourth agent group based on the final result of the first inference. Then, it further calculates the fourth evaluation metric according to the evaluation metric calculation formula.

[0073] Furthermore, the communication structure optimization system can determine whether to retain the fourth agent group based on the fourth evaluation index.

[0074] It should be noted that the second optimization process can be found in the first optimization process, and the parts that are repeated will not be described again.

[0075] It is worth noting that other reasonable combinations of steps that can be conceived by those skilled in the art based on the above description also fall within the scope of protection of this application. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to this application.

[0076] Based on the communication structure optimization method provided in the above-described embodiments, this application also provides a communication structure optimization device, which will be described below with reference to the accompanying drawings. Since the principle by which the device in this disclosure solves the problem is similar to the communication structure optimization method described above in this application, the implementation of the device can refer to the implementation of the method, and repeated details will not be elaborated further.

[0077] See Figure 5 As shown in the figure, this is a schematic diagram of a communication structure optimization device provided in an embodiment of this application. Figure 5 As shown, the communication structure optimization device 500 includes: The first determining module 501 is used to determine the first intelligent agent group, which includes multiple intelligent agents that have a cooperative relationship during the first reasoning process of the information to be processed. The second determining module 502 is used to determine the first evaluation index, which is used to indicate the collaborative effect of all agents in the first agent group. The third determining module 503 is used to determine whether to retain the first intelligent agent group based on the first evaluation index.

[0078] In one possible implementation, the device 500 further includes: The fourth determination module is used to determine the first agent in the first agent group when the first evaluation index indicates that the first agent group should be retained. The first agent is used to summarize the multiple inference results output by all agents in the first agent group to obtain the first sub-inference result, with one agent outputting one inference result.

[0079] In one possible implementation, the device 500 further includes: The sending module is used to send the first sub-inference result to the second agent. The second agent is used to summarize the multiple inference results output by all agents in the second agent group to obtain the second sub-inference result. The first agent group and the second agent group have a communication connection. The second agent group includes multiple agents that have a cooperative relationship in the first inference process of the information to be processed. The second agent is used to reason about the first sub-reasoning result and the second sub-reasoning result to obtain the first reasoning result.

[0080] In one possible implementation, the device 500 further includes: The fifth determining module is used to determine the final reasoning result of the information to be processed based on the first reasoning result and the second reasoning result. The second reasoning result is used to instruct the third agent to summarize the reasoning results of all agents in the third agent group to obtain the third sub-reasoning result. The third intelligent agent group includes multiple intelligent agents that have a cooperative relationship during the first reasoning process of the information to be processed. The third intelligent agent group has no communication connection with the first intelligent agent group and the second intelligent agent group.

[0081] In one possible implementation, the device 500 further includes: The sixth determining module is used to determine the fourth intelligent agent group, which includes multiple intelligent agents that have a cooperative relationship in the second reasoning process of the information to be processed; The fourth evaluation index is determined based on the final inference results. The fourth evaluation index is used to indicate the cooperative effect of all agents in the fourth agent group. The decision on whether to retain the fourth agent group is based on the fourth evaluation metric.

[0082] In one possible implementation, the second determining module is used to determine the first evaluation index based on the first parameter, the second parameter, the third parameter, and the fourth parameter; The first parameter indicates the number of times the first agent group successfully reasoned the information to be processed; the second parameter indicates the number of times the first agent group failed to reason the information to be processed; the third parameter indicates the evaluation score of the agent communication edge in the first agent group; the agent communication edge indicates the communication relationship between two agents; and the fourth parameter indicates the degree of duplication between the agents in the first agent group and the agents in the retained agent group.

[0083] It should be noted that the specific implementation of each module in this embodiment can be found in the relevant descriptions in the above method embodiments. The module division in this application embodiment is illustrative and only represents a logical functional division; in actual implementation, there may be other division methods. In this application embodiment, each functional module can be concentrated in one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. For example, in the above embodiment, the first acquisition module and the holding module can be the same module or different modules.

[0084] Furthermore, this application also provides a computer device, which includes a processor and a memory. The memory is used to store instructions, and the processor executes the instructions stored in the memory to cause the computer device to perform the communication structure optimization method provided in any of the above embodiments.

[0085] This application provides a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the communication structure optimization method provided in any of the above embodiments.

[0086] This application provides a computer program product containing instructions that, when run on at least one computing device, causes the at least one computing device to execute the communication structure optimization method provided in any of the above embodiments.

[0087] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems or apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to the method section.

[0088] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0089] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0090] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0091] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A communication structure optimization method, characterized in that, The method includes: A first group of intelligent agents is determined, which includes multiple intelligent agents that have a cooperative relationship during the first reasoning process of the information to be processed; A first evaluation index is determined, which is used to indicate the collaborative effect of all agents in the first agent group. Whether to retain the first intelligent agent group is determined based on the first evaluation index.

2. The method according to claim 1, characterized in that, The method further includes: If the first evaluation metric indicates that the first agent group should be retained, then the first agent group is determined. The first agent is used to summarize multiple inference results output by all agents in the first agent group to obtain the first sub-inference result, with one agent outputting one inference result.

3. The method according to claim 2, characterized in that, The method further includes: The first sub-inference result is sent to the second agent. The second agent is used to summarize the multiple inference results output by all agents in the second agent group to obtain the second sub-inference result. The first agent group and the second agent group have a communication connection. The second agent group includes multiple agents that have a cooperative relationship in the first inference process of the information to be processed. The first reasoning result is obtained by using the second agent to reason about the first sub-reasoning result and the second sub-reasoning result.

4. The method according to claim 3, characterized in that, The method further includes: Based on the first reasoning result and the second reasoning result, the final reasoning result of the information to be processed is determined. The second reasoning result is used to instruct the third agent to summarize the reasoning results of all agents in the third agent group to obtain the third sub-reasoning result. The third intelligent agent group includes multiple intelligent agents that have a cooperative relationship during the first reasoning process of the information to be processed. The third intelligent agent group has no communication connection with the first intelligent agent group and the second intelligent agent group.

5. The method according to claim 4, characterized in that, The method further includes: A fourth intelligent agent group is determined, which includes multiple intelligent agents that have a cooperative relationship during the second reasoning process of the information to be processed; A fourth evaluation index is determined based on the final reasoning result. The fourth evaluation index is used to indicate the cooperative effect of all agents in the fourth agent group. The decision on whether to retain the fourth agent group is based on the fourth evaluation metric.

6. The method according to claim 1, characterized in that, The determination of the first evaluation index includes: The first evaluation index is determined based on the first parameter, the second parameter, the third parameter, and the fourth parameter; Wherein, the first parameter is used to indicate the number of times the first agent group successfully infers the information to be processed, the second parameter is used to indicate the number of times the first agent group fails to infer the information to be processed, the third parameter is used to indicate the evaluation score of the agent communication edge in the first agent group, the agent communication edge is used to indicate that there is a communication relationship between two agents, and the fourth parameter is used to indicate the degree of duplication between the agents in the first agent group and the agents in the retained agent group.

7. A communication structure optimization device, characterized in that, The device includes: The first determining module is used to determine the first intelligent agent group, which includes multiple intelligent agents that have a cooperative relationship during the first reasoning process of the information to be processed. The second determining module is used to determine the first evaluation index, which is used to indicate the collaborative effect of all agents in the first agent group. The third determining module is used to determine whether to retain the first intelligent agent group based on the first evaluation index.

8. A computer device, characterized in that, The computer device includes a processor and a memory, the memory being used to store instructions, and the processor executing the instructions stored in the memory to cause the computer device to perform the communication structure optimization method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, It stores a computer program, wherein the computer program, when executed by a processor, implements the communication structure optimization method as described in any one of claims 1 to 6.

10. A computer program product containing instructions, characterized in that, When it is run on at least one computing device, the at least one computing device performs the communication structure optimization method as described in any one of claims 1 to 6.

Images