Graph External Attention Module for Multi-View Representation Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-view graph representation learning methods neglect external view information, leading to model inaccuracy and limited capability in learning comprehensive graph representations.
Innovation Solution
The proposed method employs a GEA-guided multi-view graph representation learning approach, utilizing external feature memory units to integrate external view information with internal graph information, and combines a graph external attention module with a message passing network and global self-attention mechanism to learn more accurate node representations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only internal information of graphs is taken into account, then the model structure remains simple, but the graph representation accuracy deteriorates due to neglecting external view information
Solution Approach 1:
The patent introduces an external view dimension by utilizing external feature memory units that store information from multiple external views (e.g., textual descriptions, images, or other data sources). The graph external attention module processes this external view information separately from the internal graph structure, adding a new dimensional aspect to the representation learning without fundamentally altering the core graph processing architecture.
Solution Approach 2:
The patent merges the processing of internal graph information and external view information by integrating their outputs through the normalization module and multilayer perceptron combination module. The external view representations and internal graph representations are combined to produce final node representations, allowing the model to leverage both internal structure and external context without requiring a completely new architectural framework.
2Adaptability or versatility
If external feature memory units are introduced to learn external view information, then the comprehensiveness of graph representations improves, but the device complexity increases
Solution Approach 1:
The external feature memory units are designed to store and process information from multiple external views, making the system versatile for different types of external data (textual, visual, tabular, etc.). The same basic architecture can handle various external view types by simply changing the input data format, without requiring fundamental redesign of the memory units or attention mechanism.
Solution Approach 2:
The graph external attention module serves as an intermediary between the external feature memory units and the final node representations. It processes external view information, attends to relevant features, and integrates this information with internal graph representations through the normalization and combination modules, acting as a mediator that bridges external and internal information sources.
Data Source
AI summary
A method and device for graph external attention (GEA)-guided multi-view graph representation learning are provided. The method includes: acquiring a node feature, an edge feature, and an adjacent matrix of an input graph; calculating, by a feature encoding module, a node embedding and an edge embedding of the graph; learning, by a global self-attention module, internal global view information of the graph, and outputting a first node representation; learning, by a message passing module, internal local view information of the graph, and outputting a second node representation and a first edge representation; learning, by a graph external attention module, external view information of the graph, and outputting a third node representation and a second edge representation; and performing, by a normalization module, batch normalization and random dropout on output information, and combining using a multilayer perceptron combination module to obtain a target output of a model.