A method and apparatus for constructing a collaboration architecture

Abstract

The application discloses a method and device for constructing a collaboration architecture, and relates to the technical field of machine learning. The method comprises the following steps: obtaining a natural language description of a federated learning requirement and a federated learning event log; obtaining a multi-node collaboration relationship based on the natural language description of the federated learning requirement through a first large language model; obtaining a collaboration workflow based on the federated learning event log through a second large language model; and obtaining a collaboration architecture of the federated learning based on the multi-node collaboration relationship and the collaboration workflow. The application utilizes the semantic understanding and analysis capabilities of the large language model to analyze the federated learning requirement and the federated learning event log, so as to construct clear multi-node collaboration relationship and collaboration workflow, thereby obtaining the collaboration architecture. Since the multi-node collaboration relationship and the collaboration workflow in the collaboration architecture are clear, efficient federated learning can be realized based on the collaboration architecture, thereby improving the efficiency of the federated learning and the performance of the model obtained by the federated learning.

Description

Technical Field

[0001] This application relates to the field of machine learning technology, and in particular to a method and apparatus for constructing a collaborative architecture. Background Technology

[0002] Federated learning is a distributed machine learning technique that allows models to be trained on multiple nodes (such as mobile phones, servers, and other devices) without directly exchanging data. Since the raw data does not need to leave the local device, federated learning can protect data privacy and security and reduce the risk of data leakage.

[0003] Furthermore, federated learning allows nodes to train the model together without sharing the original data. By combining the diversity of data from different nodes, it can help build a more comprehensive and accurate model. In addition, the distributed nature of federated learning complies with data protection regulations, which helps to enhance the compliance of data processing.

[0004] However, the multi-node collaboration relationships in current federated learning technologies are unclear, and the collaborative workflow in federated learning is complex. This makes it difficult for the multiple nodes involved in federated learning to collaborate effectively in the process of data sharing and model updates, resulting in low efficiency and poor performance of the models obtained through federated learning. Summary of the Invention

[0005] To address the aforementioned issues, this application provides a method and apparatus for constructing a collaborative architecture, which can build clear multi-node collaborative relationships and collaborative workflows, resulting in a clear collaborative architecture. This facilitates efficient federated learning based on the constructed clear collaborative architecture, thereby improving the efficiency of federated learning and the performance of the model obtained through federated learning.

[0006] The embodiments of this application disclose the following technical solutions:

[0007] Firstly, this application provides a method for constructing a collaborative architecture, including:

[0008] Obtain the federated learning requirements described in natural language and the federated learning event logs; the federated learning requirements include all technical and business requirements involved in the federated learning;

[0009] Based on the federated learning requirements described in the natural language, the multi-node collaboration relationships of the federated learning are obtained through the first large language model; the first large language model is a large language model that has received the collaboration relationship definition in advance.

[0010] The collaborative workflow of the federated learning is obtained through the second large language model based on the federated learning event log; the second large language model is a large language model that has received nested relation definitions and nested relation analysis examples in advance.

[0011] Based on the multi-node collaboration relationship and the collaboration workflow, the collaboration framework of the federated learning is obtained.

[0012] Optionally, the definition of the collaboration relationship includes: definition of constraints, definition of collaboration strategy, definition of state list, definition of state transition rules, and definition of entity permissions; the first large language model is obtained in the following way:

[0013] The constraints, collaboration strategies, state lists, state transition rules, and entity permissions are each input into the large language model in their corresponding templates to obtain the first large language model.

[0014] Optionally, the step of obtaining the multi-node collaborative relationship of the federated learning based on the federated learning requirements described by the natural language using the first large language model includes:

[0015] The federated learning requirements described in the natural language are input into the first large language model so that the first large language model can analyze the federated learning requirements described in the natural language to obtain the target information in the federated learning requirements; and based on the target information and the templates corresponding to the definition of the constraint conditions, the definition of the collaboration strategy, the definition of the state list, the definition of the state transition and the definition of the entity permission, the multi-node collaboration relationship of the federated learning is obtained and output; the target information is information related to the definition of the collaboration relationship.

[0016] Optionally, the second largest language model is obtained in the following way:

[0017] The nesting relationship definition is input into the large language model; wherein, the nesting relationship definition includes the concept and representation of the nesting relationship, and the concept of the nesting relationship is that the nesting relationship represents the calling relationship between the parent procedure and the child procedure;

[0018] By inputting federated learning analysis examples into the large language model, a second large language model is obtained; wherein, the federated learning analysis examples include: event log examples and nested relationship examples corresponding to the event log examples.

[0019] Optionally, when the second large language model is a multimodal large language model, the federated learning analysis example further includes: a nested relation tree example corresponding to the event log example.

[0020] Optionally, the step of obtaining the collaborative workflow of the federated learning based on the federated learning event log through the second major language model includes:

[0021] The federated learning event log is input into the second large language model so that the second large language model can analyze the federated learning event log, obtain and output the nesting relationship corresponding to the federated learning event log;

[0022] Based on the nesting relationship of the federated learning event logs, the collaborative workflow corresponding to the federated learning is obtained.

[0023] Optionally, the method further includes: obtaining the target federated learning event, wherein the target federated learning event is a federated learning event that requires collaborative workflow sorting;

[0024] The step of inputting the federated learning event log into the second large language model, so that the second large language model can analyze the federated learning event log, obtain and output the nesting relationship corresponding to the federated learning event log, includes:

[0025] The federated learning event log and the target federated learning event are input into the second large language model so that the second large language model can analyze the target federated learning event in the federated learning event log, obtain and output the nesting relationship between the target federated learning events;

[0026] The process of arranging the nested relationships corresponding to the federated learning event logs to obtain the collaborative workflow corresponding to the federated learning includes:

[0027] Based on the nesting relationships between the target federated learning events, the collaborative workflow corresponding to the federated learning is arranged.

[0028] Optionally, the federated learning event log is obtained in the following way:

[0029] Retrieve local event logs from multiple nodes in federated learning;

[0030] The events in the local event logs of the multiple nodes are sorted in chronological order to obtain the corresponding federated learning event logs.

[0031] Secondly, this application provides a device for constructing a collaborative architecture, characterized in that it includes:

[0032] The federated data acquisition module is used to acquire federated learning requirements described in natural language and federated learning event logs; the federated learning requirements include all technical and business requirements involved in the federated learning.

[0033] The collaboration relationship construction module is used to obtain the multi-node collaboration relationship of the federated learning based on the federated learning requirements described by the natural language through the first large language model; the first large language model is a large language model that has received the collaboration relationship definition in advance;

[0034] The collaborative workflow construction module is used to obtain the collaborative workflow of the federated learning based on the federated learning event log through the second large language model; the second large language model is a large language model that has received nested relation definitions and nested relation analysis examples in advance;

[0035] The collaboration architecture building module is used to obtain the collaboration framework of the federated learning based on the multi-node collaboration relationship and the collaboration workflow.

[0036] Optionally, the collaboration relationship construction module is specifically used to input the federated learning requirements described in natural language into the first language model, so that the first language model can analyze the federated learning requirements described in natural language and obtain the target information in the federated learning requirements; and based on the target information, as well as the templates corresponding to the definitions of constraints, collaboration strategies, state lists, state transitions and entity permissions, obtain and output the multi-node collaboration relationship of the federated learning.

[0037] Optionally, a collaborative workflow building module is used to input federated learning event logs into the second language model so that the second language model can analyze the federated learning event logs, obtain and output the nesting relationship corresponding to the federated learning event logs; and arrange the collaborative workflow corresponding to the federated learning based on the nesting relationship corresponding to the federated learning event logs.

[0038] Optionally, the federated data acquisition module is also used to acquire the target federated learning events;

[0039] The collaborative workflow construction module is specifically used to input federated learning event logs and target federated learning events into the second language model, so that the second language model can analyze the target federated learning events in the federated learning event logs, obtain and output the nesting relationship between the target federated learning events; based on the nesting relationship between the target federated learning events, the collaborative workflow corresponding to the federated learning is arranged.

[0040] Compared to existing technologies, this application offers the following advantages: It leverages the semantic understanding and analysis capabilities of a large language model to parse / analyze federated learning requirements and event logs, thereby constructing a clear multi-node collaborative relationship and workflow, resulting in a collaborative architecture for federated learning. Because the multi-node collaborative relationship and workflow within the collaborative architecture are clear, efficient federated learning can be achieved based on this architecture, thus improving the efficiency of federated learning and the performance of the model obtained through federated learning. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 A flowchart illustrating a method for constructing a collaborative architecture as provided in an embodiment of this application;

[0043] Figure 2 Example diagram of an event log provided in an embodiment of this application;

[0044] Figure 3 An example diagram of a nested relationship tree provided in an embodiment of this application;

[0045] Figure 4 Example diagram of a collaborative workflow provided in the embodiments of this application;

[0046] Figure 5 This is a schematic diagram of a construction device for a collaborative architecture provided in an embodiment of this application. Detailed Implementation

[0047] In federated learning, multi-node collaboration refers to the cooperative model among different nodes participating in the federated learning process. Multi-node collaboration in federated learning ensures that different nodes can work effectively together to achieve a common goal: training a high-quality machine learning model. However, due to trust issues between nodes and the heterogeneity of data across nodes, federated learning faces the problem of complex multi-node collaboration. The collaborative workflow refers to how multiple nodes collaborate to complete the federated learning process. However, federated learning involves multiple stages such as data isolation between nodes, model training, and model aggregation, leading to a complex collaborative workflow.

[0048] In current technologies, the multi-node collaboration relationships involved in federated learning are unclear, and the collaborative workflow in federated learning is complex. This makes it difficult for multiple nodes involved in federated learning to effectively coordinate during data sharing and model updates, resulting in low efficiency and poor performance of the models obtained through federated learning.

[0049] Furthermore, due to the low efficiency of federated learning and the poor performance of the models obtained from it, federated learning is difficult to be widely used in practice (i.e., its application penetration rate is low), thus limiting its potential in protecting data privacy and cross-node collaboration.

[0050] This application provides a method for constructing a collaborative architecture, including: obtaining federated learning requirements described in natural language and federated learning event logs; obtaining multi-node collaborative relationships in federated learning based on the federated learning requirements described in natural language using a first large language model; obtaining the collaborative workflow of federated learning based on the federated learning event logs using a second large language model; and obtaining the collaborative architecture of federated learning based on the multi-node collaborative relationships and collaborative workflows. This application utilizes the semantic understanding and analysis capabilities of large language models to parse / analyze the federated learning requirements and federated learning event logs to construct clear multi-node collaborative relationships and collaborative workflows, thereby obtaining the collaborative architecture of federated learning. Because the multi-node collaborative relationships and collaborative workflows in the collaborative architecture are clear, efficient federated learning can be achieved based on the collaborative architecture, thereby improving the efficiency of federated learning and the performance of the model obtained through federated learning.

[0051] Furthermore, the improved efficiency and performance of federated learning models can increase its adoption in practice, thereby maximizing its ability to protect data privacy and facilitate cross-node collaboration.

[0052] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0053] Example 1:

[0054] The following is combined with Figures 1-4 This application provides a detailed description of a method for constructing a collaborative architecture.

[0055] like Figure 1 As shown in the embodiments of this application, a method for constructing a collaborative architecture includes the following steps:

[0056] S101. Obtain the federated learning requirements for natural language descriptions and the federated learning event log.

[0057] Among them, federated learning requirements described in natural language refer to the use of natural language (such as Chinese, English, etc.) to express federated learning requirements.

[0058] Federated learning requirements encompass all the technical and business requirements involved. Specifically, the technical requirements include: data requirements (such as data format, data volume, and data quality), model requirements (such as model type and model performance), training requirements (such as training algorithms and training frequency), communication requirements (such as communication protocols and communication frequencies), and computing resources (such as hardware requirements and storage requirements). Business requirements include: business objectives, nodes (i.e., participants), incentive mechanisms, cooperation models, and compliance.

[0059] The federated learning event log is a time-series data record of various federated learning events that occur during the federated learning process. It typically contains key information about the execution of the federated learning process, such as the start, completion, and status changes of federated learning events. For example, the federated learning event log includes: event name, event status, and timestamp.

[0060] In one possible implementation, the federated learning event log is obtained as follows: the local event logs of multiple nodes in the federated learning are obtained; the times in the local time logs of the multiple nodes are sorted in chronological order to obtain the corresponding federated learning event log.

[0061] Specifically, after obtaining the local event logs of multiple nodes in the federated learning process, the local event logs of all nodes are merged into a list or data structure; all events are sorted according to the timestamp order to ensure that the events are arranged in chronological order, thus obtaining the corresponding federated learning event logs.

[0062] S102. Using the first major language model, based on the federated learning requirements described by natural language, the multi-node collaborative relationship of federated learning is obtained.

[0063] The first major language model is a large language model that receives pre-defined collaborative relationships.

[0064] Large Language Models (LLMs) are large-scale pre-trained language models that generate high-quality natural language text by training on large amounts of text data. Beyond generating natural language text, LLMs can capture complex patterns and semantic structures in language, gain a deep understanding of the meaning of language text, and handle various language tasks.

[0065] Specifically, the first large language model is obtained by inputting the definition of the collaboration relationship into the large language model.

[0066] The definition of collaborative relationships includes: definitions of constraints, collaboration strategies, state lists, state transition rules, and entity permissions. These definitions are then input into the large language model to obtain the first large language model.

[0067] The constraint definition is used to define the name and format of the content included in the constraint; the collaboration policy definition is used to define the name and format of the content included in the collaboration policy; the state list definition is used to define the name and format of the content included in the state list; the state transition rule definition is used to define the name and format of the content included in the state transition rule; and the entity permission definition is used to define the name and format of the content included in the entity permission.

[0068] In one possible implementation, the definitions of restricted conditions, collaboration strategies, state lists, state transition rules, and entity permissions are input into the large language model in the form of their corresponding templates to obtain the first large language model.

[0069] To facilitate understanding, the templates corresponding to the definition of collaboration relationships are illustrated below with examples from Tables 1-5. The templates for collaboration relationships are explained using tables as an example.

[0070] The following example, based on Table 1, illustrates the templates corresponding to the definition of restricted conditions.

[0071] Table 1

[0072]

[0073] The templates for defining the constraints are shown in Table 1. The constraints include: list of initiators, list of initiating groups, list of completers, list of completers, maximum number of rounds, and rejection policy; as well as the format types of various constraints.

[0074] The initiator list includes nodes (participants) responsible for launching federated learning tasks; the initiator group list includes groups of nodes responsible for launching federated learning tasks. Specifically, only nodes in the initiator list or whose groups belong to groups in the initiator group list can launch federated learning tasks.

[0075] The list of completers includes nodes (participants) responsible for completing the federated learning task; the list of completegroups includes groups of nodes that are responsible for completing the federated learning task. Specifically, only nodes in the completer list or whose groups belong to groups in the completegroup list can complete the federated learning task.

[0076] The initiator list, initiator group list, completer list, and completer group list above are formatted as JSON arrays. JSON (JavaScript Object Notation) is a lightweight data-interchange format that is easy for humans to read and write, and also easy for machines to parse and generate. A JSON-formatted array is an ordered collection of values, typically enclosed in square brackets and separated by commas. For example, "initiator": ["init_user1", "init_user2"].

[0077] The maximum number of rotations (maxTurns) represents the maximum upper limit on the number of times the federated learning task is passed between nodes during the entire training process. Specifically, when the maximum number of rotations is -1, it means there is no limit to the number of times the federated learning task is passed.

[0078] The rejection strategy represents the action triggered when the federated learning task reaches a certain boundary condition. This boundary condition can be exceeding the maximum number of iterations, reaching a deadline, or other preset conditions. Specifically, there are three rejection strategies: WARNING, EXCEPTION, and IGNORE. WARNING means issuing a warning message when the federated learning task reaches a certain boundary condition but continuing execution; EXCEPTION means issuing an exception message and stopping execution when the federated learning task reaches a certain boundary condition; IGNORE means ignoring the boundary condition and making no record when the federated learning task reaches a certain boundary condition. The rejection strategy is enumerated, meaning it can be any one of the three options: WARNING, EXCEPTION, or IGNORE. For example, "rejectStrategy": "WARNING".

[0079] The following example, based on Table 2, illustrates the templates corresponding to protocol policy definitions.

[0080] Table 2

[0081]

[0082] The templates corresponding to the collaboration strategy definition are shown in Table 2. The collaboration strategy definition includes: source user / group, whether to configure group level, delivery method, target collaborator list, target collaboration group list, target adaptation method, intra-group delivery method, message notification configuration, and various collaboration strategy format types.

[0083] The source user / group (SourceId) includes the ID of the node / node group in the current federated learning task where the collaboration policy is effective. Specifically, the collaboration policy is effective when the ID of the current collaborating node or node group in the federated learning task is the ID in the source user / group.

[0084] The "IsGroup-Level Configuration" option indicates whether the collaboration policy applies to groups. This configuration is a Boolean value, which can only take one of two values: True or False. Specifically, when the "IsGroup-Level Configuration" is True, the protocol policy applies to the entire group; when it is False, the policy applies only to a single node.

[0085] The transmission method (transmitType) indicates how the federated learning task or model parameters are transmitted between nodes. Transmission methods include: DELIVER, CLAIM, and ASSIGN. DELIVER means that the initiator directly sends the task or model parameters to the designated participant; CLAIM means that the participant actively requests the task or model parameters from the initiator; and ASSIGN means that the initiator sends the task or model parameters to the participant according to a specific adaptation method.

[0086] It should be noted that the initiator refers to the node that initiates the federated learning task, such as the scheduler or central processor of the federated learning task; the participant refers to the node that participates in the federated learning task, such as the client that has a local dataset.

[0087] The target collaborator list (targetCoopers) includes nodes (participants) that can receive federated learning tasks or model parameters; the target collaborator group list (targetCoopGroups) includes groups of nodes (participant groups) that can receive federated learning tasks or model parameters. Specifically, in the above transmission method, the transmission target (participant) should be in the target collaborator list, or the group to which the transmission target (participant) belongs should be in the target collaborator group list.

[0088] The target adaptation method (adaptType) indicates the adaptation method used when dispatching tasks or models to participants. Specifically, the initiator sends task or model parameters to participants according to the target adaptation method. Target adaptation methods include: engine-built-in random adaptation (random) and optimization algorithm adaptation (betterFit). Engine-built-in random adaptation means the initiator randomly sends task or model parameters to participants; optimization algorithm adaptation means the initiator sends task or model parameters to suitable participants according to an optimization algorithm.

[0089] The innerTransmitType is used to indicate how federated learning tasks or model parameters are transmitted within a group when the group-level configuration is in effect.

[0090] Among them, the message notification configuration (notice) refers to the notification information triggered after the task or model parameters are passed.

[0091] The following example, using Table 3 as a reference, illustrates the template corresponding to the definition of a status list.

[0092] Table 3

[0093]

[0094] The template for the state list definition is shown in Table 3. The state list includes: initial state of the entity, final state of the entity, state value, and state description. The state value and state description can be referred to as the state unit configuration. In federated learning, an entity represents a node in the federated learning process.

[0095] The initial status of an entity represents the initial state of a node when the federated learning task is first initiated.

[0096] The final status of an entity represents the final state of a node after the federated learning task is completed.

[0097] The status unit configuration represents a single status information of an entity, including a status value (status) and a status description (text). The status value (status) represents a unique identifier for the status, such as an identifier or code; the status description (text) represents a detailed textual description of the status, which is also the information displayed on the page.

[0098] It should be noted that, generally speaking, an entity will be configured with multiple state unit configurations.

[0099] The templates corresponding to the state transition rule definitions are described in detail below with reference to Table 4.

[0100] Table 4

[0101]

[0102] The templates for defining state transition rules are shown in Table 4. The state transition rules are defined as follows: current state, transition action, and target state.

[0103] Here, the current status represents the current state of the entity; the action represents the action / operation that triggers the state transition, that is, the action can trigger the transition from the current state to the target state; and the target status represents the new state of the entity after the state transition.

[0104] Specifically, when a federated learning task or model parameters are passed, state transitions occur among the associated entities. The state transition rules for each entity are interconnected to form a finite state machine. Furthermore, the current state and the target state in the state transition rules need to be pre-set in the state list.

[0105] The templates corresponding to entity permission definitions are described in detail below with reference to Table 5.

[0106] Table 5

[0107]

[0108] The templates corresponding to entity permission definitions are shown in Table 5. The entity permissions are defined as follows: source user / group, whether to group level configuration and permission value.

[0109] The source user / group (SourceId) includes the ID of the node / node group whose entity permissions are effective. Specifically, the entity permissions are effective when the ID of the current collaborating node or node group in the federated learning task is the ID in the source user / group.

[0110] The "isGroup" option indicates whether the permission definition applies to groups.

[0111] The permission value represents the entity permissions that the source user / group possesses.

[0112] Specifically, the permission value format is "Cooperation Time: Operation Permission". The collaboration time includes before, during, and after collaboration. Before collaboration refers to a node not yet joining the collaboration chain of the federated learning task, and its permission value is represented as `beforeCoop`. During collaboration, the node is the current executing node of the federated learning task, and its permission value is represented as `cooping`. After collaboration, the node joins the collaboration chain after passing on the federated learning task, and its permission value is represented as `afterCoop`. Operation permissions include: read permission, write permission, execute permission, and delete permission. For example, having write permission during collaboration corresponds to the permission value "cooping: write".

[0113] The above examples, using Tables 1-5, illustrate the templates corresponding to the collaborative relationship definitions. For instance, the first large language model is obtained by inputting the templates corresponding to the collaborative relationship definitions shown in Tables 1-5 into the large language model.

[0114] In one possible implementation, the federated learning requirements described in natural language are input into the first large language model; the first large language model analyzes the federated learning requirements described in natural language to obtain the target information in the federated learning requirements; based on the target information and the templates corresponding to the definitions of constraints, collaboration strategies, state lists, state transitions, and entity permissions, the first large language model constructs and outputs the multi-node collaboration relationship of the federated learning to obtain the multi-node collaboration relationship of the federated learning.

[0115] This application utilizes the semantic understanding and analysis capabilities of a large language model to deeply analyze federated learning requirements and construct accurate and clear collaborative relationships based on pre-input collaborative relationship definitions, thereby clearly representing the complex collaborative relationships between federated learning nodes.

[0116] S103. Using the second language model, based on the federated learning event log, obtain the collaborative workflow of federated learning.

[0117] The second largest language model is a large language model that receives nested relation definitions and a full set of relation analysis examples in advance.

[0118] Specifically, the second large language model is obtained by inputting the nested relation definition into the large language model and inputting federated learning analysis examples into the large language model to obtain the second large language model.

[0119] The definition of nested relationships includes the concept and representation of nested relationships. Defining nested relationships for large language models is used to define / inform the large language model what nested relationships are and how they should be represented.

[0120] Nested relationships are defined as the calling relationship between parent and child processes. In other words, a nested relationship means that a process (parent process) can contain or call other processes (child processes), forming a hierarchical structure. For example, after inputting the nested relationship definition into a large language model, the model can determine that if, in the event log, the lifecycle of a process from "Start" to "Complete" includes the complete trajectories of one or more other processes, then a nested relationship exists between the process (parent process) and other processes (child processes).

[0121] Representation of nested relationships: If event a is nested within event b, that is, event a is the parent process and event b is the child process, then it is represented as a<-b.

[0122] The federated learning analysis examples include event log examples and examples of nested relationships between them. Inputting these examples into the large language model is to define / inform how to apply these federated learning definitions in specific situations. Furthermore, the federated learning analysis examples are derived from real federated learning projects or virtual datasets, demonstrating how to obtain the nested relationships between events from the federated learning event logs.

[0123] To make it easier to understand, the following will be combined with... Figure 2 Here are some examples of event logs and their nested relationships.

[0124] like Figure 2 The example event log shown includes event identifiers, such as event name and event number; the lifecycle includes two types: event start (Start) and event end (Complete); and timestamps record the specific time when each event occurred, such as the timestamp corresponding to the start of the event and the timestamp corresponding to the end of the event.

[0125] like Figure 2 The example event log shown includes eight events: Event A, Event B, Event C, Event D, Event E, Event F, Event G, and Event H. The order in the event log example is based on the timestamps corresponding to the lifecycles of each event, sorted chronologically. For example, the time domain of Event B's lifecycle "Start" to "Complete" includes the lifecycles of Event E and Event F, so Event B nests Events E and F, denoted as B <- E; B <- F.

[0126] and Figure 2 The nesting relationships shown in the time log example are: B<-E; B<-F; C<-G; C<-H. That is, event B nests event E; event B nests event F; event C nests event G; and event C nests event H.

[0127] In one possible implementation, when the second largest language model is a multimodal large language model, the federated learning analysis example also includes: the nested relation tree example corresponding to the event log example.

[0128] Among them, the multimodal large language model is a large-scale pre-trained model capable of processing various types of data. It can understand and generate information in multiple modalities such as text, images, and audio. Therefore, for the multimodal large language model, a nested relation tree example corresponding to an event log example can be input so that the second large language model can output a nested relation tree corresponding to the federated learning log.

[0129] To make it easier to understand, the following will be combined with... Figure 3 Here's an example of a nested relationship tree. Figure 3 The nested relationship tree example shown is related to Figure 2 The example shown is a nested tree corresponding to the event log example.

[0130] like Figure 3 As shown, the nested relationship tree example represents event B nested with events E and F, and event C nested with events G and H.

[0131] Furthermore, such as Figure 3 The nested relationship tree example shown can be represented by formula (1).

[0132] Tree(W)=root, Node(B), Node(C), f (1)

[0133] Where W represents as Figure 2 The example event log shown; root represents the set of root nodes, Node represents the set of child nodes of the root node; f is the mapping from the root node to its child node set, thus forming a tree structure. Specifically, root = {A, B, C, D}, Node(B) = {E, F}, Node(C) = {G, H}.

[0134] In one possible implementation, the federated learning event log is input into the second language model; the second language model analyzes the federated learning event log to obtain and output the nesting relationship corresponding to the federated learning event log; based on the nesting relationship, the nesting relationship corresponding to the learning log is arranged to obtain the collaborative workflow corresponding to the federated learning.

[0135] Furthermore, when the second largest language model is a multimodal large language model, the second largest language model can analyze the federated learning event logs and obtain and output the nested relation tree corresponding to the federated learning event logs; based on the nested relation tree corresponding to the nested relation learning logs, the collaborative workflow corresponding to the federated learning can be obtained.

[0136] In one possible implementation, while acquiring the federated learning requirements described in natural language and the federated learning event log, the target federated learning events can also be obtained. These target federated learning events are those requiring collaborative workflow sequencing. The federated learning event log and the target federated learning events are then input into a second language model. The second language model analyzes the target federated learning events in the federated learning event log, obtaining and outputting the nesting relationships between them. This facilitates the subsequent generation of a collaborative workflow that includes only the target federated learning events.

[0137] It should be noted that when the target federated learning event is not obtained, all federated learning events in the federated learning event log by default need to be sorted according to the collaborative workflow.

[0138] To make it easier to understand, the following will be combined with... Figure 4 Examples (a) and (b) illustrate the resulting collaborative workflow, both based on... Figure 2 The nested relationship obtained from the event logs shown is as follows: Figure 3 The nested relationship tree shown represents the collaborative workflow obtained.

[0139] like Figure 4 The collaborative workflow shown in (a) is for when the target federated learning event has not been obtained, meaning that by default, all federated learning events in the federated learning event log need to be sorted in the collaborative workflow. In this case, as shown in (a), the collaborative workflow is as follows: Figure 4 The collaborative workflow shown in (a) includes all federated learning events in the event log.

[0140] like Figure 4 The collaborative workflow shown in (b) obtains target federated learning events A, C, D, G and H. The resulting collaborative workflow only includes the target federated learning events in the event log.

[0141] This application's embodiments utilize the language understanding and task decomposition capabilities of a large language model to parse the federated learning event logs, revealing clear nested relationships between events and thus a clear collaborative workflow. While obtaining a clear collaborative workflow, the complexity of the workflow is also reduced.

[0142] S104. Based on multi-node collaboration relationships and collaborative workflows, a collaborative architecture for federated learning is obtained.

[0143] Specifically, the collaborative architecture of federated learning can guide the federated learning system (i.e., each node) to perform efficient federated learning, thereby improving the efficiency of federated learning and the performance of the model obtained by federated learning.

[0144] This application provides a method for constructing a collaborative architecture, including: obtaining federated learning requirements described in natural language and federated learning event logs; obtaining multi-node collaborative relationships in federated learning based on the federated learning requirements described in natural language using a first large language model; obtaining the collaborative workflow of federated learning based on the federated learning event logs using a second large language model; and obtaining the collaborative architecture of federated learning based on the multi-node collaborative relationships and collaborative workflows. This application utilizes the semantic understanding and analysis capabilities of large language models to parse / analyze the federated learning requirements and federated learning event logs to construct clear multi-node collaborative relationships and collaborative workflows, thereby obtaining the collaborative architecture of federated learning. Because the multi-node collaborative relationships and collaborative workflows in the collaborative architecture are clear, efficient federated learning can be achieved based on the collaborative architecture, thereby improving the efficiency of federated learning and the performance of the model obtained through federated learning.

[0145] Furthermore, the improved efficiency and performance of federated learning models can increase its adoption in practice, thereby maximizing its ability to protect data privacy and facilitate cross-node collaboration.

[0146] Example 2:

[0147] The following is combined with Figure 5 This application provides a detailed description of a collaborative architecture construction apparatus.

[0148] like Figure 5 As shown in the embodiment of this application, a collaborative architecture construction apparatus includes the following modules:

[0149] The federated data acquisition module 501 is used to acquire federated learning requirements described in natural language and federated learning event logs; the federated learning requirements include all technical and business requirements involved in federated learning.

[0150] The collaboration relationship construction module 502 is used to obtain the multi-node collaboration relationship of federated learning based on the federated learning requirements described by natural language through the first language model;

[0151] Collaborative workflow building module 503 is used to obtain the collaborative workflow of federated learning based on the federated learning event log through the second language model;

[0152] Collaboration architecture building module 504 is used to obtain a collaborative framework for federated learning based on multi-node collaboration relationships and collaborative workflows.

[0153] In one possible implementation, the collaboration relationship construction module 502 is specifically used to input the federated learning requirements described in natural language into the first large language model, so that the first large language model can analyze the federated learning requirements described in natural language and obtain the target information in the federated learning requirements; and based on the target information, as well as the templates corresponding to the definitions of constraints, collaboration strategies, state lists, state transitions and entity permissions, obtain and output the multi-node collaboration relationship of the federated learning.

[0154] In one possible implementation, the collaborative workflow construction module 503 is specifically used to input the federated learning event log into the second language model so that the second language model can analyze the federated learning event log, obtain and output the nesting relationship corresponding to the federated learning event log; and arrange the collaborative workflow corresponding to the federated learning based on the nesting relationship corresponding to the federated learning event log.

[0155] In one possible implementation, the federated data acquisition module 501 is also used to acquire the target federated learning events; the collaborative workflow construction module 503 is specifically used to input the federated learning event log and the target federated learning events into the second language model, so that the second language model can analyze the target federated learning events in the federated learning event log, obtain and output the nesting relationship between the target federated learning events; and arrange the collaborative workflow corresponding to the federated learning based on the nesting relationship between the target federated learning events.

[0156] This application provides a collaborative architecture construction apparatus, including: a federated data acquisition module 501, used to acquire federated learning requirements described in natural language and federated learning event logs; the federated learning requirements include all technical and business requirements involved in federated learning; a collaborative relationship construction module 502, used to obtain multi-node collaborative relationships for federated learning based on the federated learning requirements described in natural language using a first large language model; a collaborative workflow construction module 503, used to obtain the collaborative workflow for federated learning based on the federated learning event logs using a second large language model; and a collaborative architecture construction module 504, used to obtain the collaborative framework for federated learning based on the multi-node collaborative relationships and collaborative workflows. This application utilizes the semantic understanding and analysis capabilities of large language models to parse / analyze the federated learning requirements and federated learning event logs to construct clear multi-node collaborative relationships and collaborative workflows, thereby obtaining a collaborative architecture for federated learning. Because the multi-node collaborative relationships and collaborative workflows in the collaborative architecture are clear, efficient federated learning can be achieved based on the collaborative architecture, thereby improving the efficiency of federated learning and the performance of the model obtained from federated learning.

[0157] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the method and apparatus embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiments. The method and apparatus embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components indicated as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the solution in this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0158] The above description is merely one specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for constructing a collaborative architecture, characterized in that, include: Obtain federated learning requirements and federated learning event logs in natural language description; The federated learning requirements include all the technical and business requirements involved in the federated learning process; Based on the federated learning requirements described in the natural language, the multi-node collaboration relationships of the federated learning are obtained through the first large language model; the first large language model is a large language model that has received the collaboration relationship definition in advance. The collaborative workflow of the federated learning is obtained through the second large language model based on the federated learning event log; the second large language model is a large language model that has received nested relation definitions and nested relation analysis examples in advance. Based on the multi-node collaboration relationship and the collaboration workflow, the collaboration framework of the federated learning is obtained; The definition of the collaboration relationship includes: definition of constraints, definition of collaboration strategy, definition of state list, definition of state transition rules, and definition of entity permissions; the first large language model is obtained in the following way: The constraints, collaboration strategies, state lists, state transition rules, and entity permissions are each input into the large language model in their corresponding templates to obtain the first large language model. The second major language model was obtained in the following way: The nesting relationship definition is input into the large language model; wherein, the nesting relationship definition includes the concept and representation of the nesting relationship, and the concept of the nesting relationship is that the nesting relationship represents the calling relationship between the parent procedure and the child procedure; By inputting federated learning analysis examples into the large language model, a second large language model is obtained; wherein, the federated learning analysis examples include: event log examples and nested relationship examples corresponding to the event log examples.

2. The method according to claim 1, characterized in that, The process involves using the first large language model to obtain the multi-node collaborative relationships of the federated learning based on the natural language description's federated learning requirements, including: The federated learning requirements described in the natural language are input into the first large language model so that the first large language model can analyze the federated learning requirements described in the natural language to obtain the target information in the federated learning requirements; and based on the target information and the templates corresponding to the definition of the constraint conditions, the definition of the collaboration strategy, the definition of the state list, the definition of the state transition rule, and the definition of the entity permission, the multi-node collaboration relationship of the federated learning is obtained and output; the target information is information related to the definition of the collaboration relationship.

3. The method according to claim 1, characterized in that, When the second large language model is a multimodal large language model, the federated learning analysis example also includes: the nested relation tree example corresponding to the event log example.

4. The method according to claim 1, characterized in that, The collaborative workflow of the federated learning, obtained through the second major language model and based on the federated learning event log, includes: The federated learning event log is input into the second large language model so that the second large language model can analyze the federated learning event log, obtain and output the nesting relationship corresponding to the federated learning event log; Based on the nesting relationship of the federated learning event logs, the collaborative workflow corresponding to the federated learning is obtained.

5. The method according to claim 4, characterized in that, The method further includes: Obtain the target federated learning event, which is a federated learning event that requires collaborative workflow sorting; The step of inputting the federated learning event log into the second large language model, so that the second large language model can analyze the federated learning event log, obtain and output the nesting relationship corresponding to the federated learning event log, includes: The federated learning event log and the target federated learning event are input into the second large language model so that the second large language model can analyze the target federated learning event in the federated learning event log, obtain and output the nesting relationship between the target federated learning events; The process of arranging the nested relationships corresponding to the federated learning event logs to obtain the collaborative workflow corresponding to the federated learning includes: Based on the nesting relationships between the target federated learning events, the collaborative workflow corresponding to the federated learning is arranged.

6. The method according to claim 1, characterized in that, The federated learning event log is obtained in the following way: Retrieve local event logs from multiple nodes in federated learning; The events in the local event logs of the multiple nodes are sorted in chronological order to obtain the corresponding federated learning event logs.

7. A device for constructing a collaborative architecture, characterized in that, include: The federated data acquisition module is used to acquire federated learning requirements described in natural language and federated learning event logs; the federated learning requirements include all technical and business requirements involved in the federated learning. The collaboration relationship construction module is used to obtain the multi-node collaboration relationship of the federated learning based on the federated learning requirements described by the natural language through the first large language model; the first large language model is a large language model that has received the collaboration relationship definition in advance; The collaborative workflow construction module is used to obtain the collaborative workflow of the federated learning based on the federated learning event log through the second large language model; the second large language model is a large language model that has received nested relation definitions and nested relation analysis examples in advance; A collaborative architecture building module is used to obtain the collaborative framework of the federated learning based on the multi-node collaborative relationship and the collaborative workflow. The definition of the collaboration relationship includes: definition of constraints, definition of collaboration strategy, definition of state list, definition of state transition rules, and definition of entity permissions; the first large language model is obtained in the following way: The constraints, collaboration strategies, state lists, state transition rules, and entity permissions are each input into the large language model in their corresponding templates to obtain the first large language model. The second major language model was obtained in the following way: The nesting relationship definition is input into the large language model; wherein, the nesting relationship definition includes the concept and representation of the nesting relationship, and the concept of the nesting relationship is that the nesting relationship represents the calling relationship between the parent procedure and the child procedure; By inputting federated learning analysis examples into the large language model, a second large language model is obtained; wherein, the federated learning analysis examples include: event log examples and nested relationship examples corresponding to the event log examples.

Images