A task-driven agent dynamic decision optimization method and system

By introducing a task-driven dynamic decision-making mechanism into the graph neural network and optimizing the node receptive field structure, the problem of inaccurate information aggregation caused by static receptive fields is solved, improving the accuracy and generalization of the graph representation learning model, especially in the prediction of social networks and protein interactions.

CN122154754APending Publication Date: 2026-06-05XINJIANG TECH INST OF PHYSICS & CHEM CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Applications(China)
Current Assignee / Owner
XINJIANG TECH INST OF PHYSICS & CHEM CHINESE ACAD OF SCI
Filing Date
2026-02-02
Publication Date
2026-06-05

AI Technical Summary

Technical Problem

Existing technologies using static receptive fields in graph neural networks lead to inaccurate information aggregation, making it difficult to dynamically adjust based on task objectives and node positions. This introduces redundant information or ignores key structural relationships, affecting the accuracy and generalization of task decisions.

Method used

By constructing neighborhood agents and layer agents, reinforcement learning is used to optimize the node receptive field structure, dynamically determine the optimal neighbor selection strategy and network layer number, and combine attention mechanism and self-supervised loss function to optimize node embedding and generate task-specific node representations.

Benefits of technology

It effectively reduces redundant information interference, highlights key local structural relationships, and improves the performance of graph representation learning models in downstream tasks, especially in the accuracy and computational efficiency of large-scale social network analysis and protein interaction prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122154754A_ABST
    Figure CN122154754A_ABST
Patent Text Reader

Abstract

The application discloses a task-driven intelligent agent dynamic decision optimization method and system, which comprises the following steps: modeling graph structure data, constructing an adjacency matrix, a feature matrix and a similarity matrix; constructing neighborhood intelligent agents and layer number intelligent agents; based on the neighborhood intelligent agents, evaluating the importance of neighbor nodes through an attention mechanism and combining the upper confidence bound method to make reinforcement learning decisions, so as to obtain an optimal neighbor selection strategy; based on the layer number intelligent agents, dynamically determining the optimal aggregation network layer number for each node to obtain an optimal network layer number identification strategy; based on the above two strategies, re-normalizing the attention weight of the node and executing task-specific information aggregation under a graph neural network framework to generate a node embedding representation; constructing a joint loss function comprising a reconstruction loss and a self-supervised loss, optimizing the joint loss function, training the parameters of the neighborhood intelligent agents, the layer number intelligent agents and the graph neural network, and outputting an optimal node embedding result.
Need to check novelty before this filing date? Find Prior Art