Temporal Co-Graph Networks for Multi-Type Node Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Temporal Graph Networks (TGNs) are limited in handling multiple types of nodes and cannot model interactions between nodes of the same type, restricting their ability to handle complex relationships in graphs with diverse entities.
Innovation Solution
The solution extends TGNs by allowing multiple TGN graphs to interact, enabling nodes from one graph to inform nodes in another graph, and sharing memory state vector mappings, which facilitates concurrent training and synchronization of edge features across graphs, avoiding memory allocation issues through dynamic generation of edge features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If TGN is extended to handle multiple types of nodes with interactions within and across types, then the capability to model complex graph structures is improved, but the device complexity and memory requirements increase
Solution Approach 1:
The patent divides the complex multi-type node graph into multiple separate TGN graphs, where each TGN graph handles a specific node type. This segmentation allows each sub-graph to be processed independently while maintaining the overall complexity management through modular architecture.
Solution Approach 2:
The patent creates a universal framework where multiple TGN graphs can interact through shared memory mechanisms. The memory buffer serves multiple functions: storing node features, enabling cross-graph interactions, and facilitating information exchange between different node types, thus providing multi-functionality that handles diverse graph structures.
2Adaptability or versatility
If multiple TGN graphs are introduced to handle different node types, then the versatility of the model is improved, but the memory allocation and management complexity worsen
Solution Approach 1:
The patent merges multiple TGN graphs into a unified system through a shared memory buffer. Instead of maintaining separate memory allocations for each graph type, the system combines them into a single memory structure that all TGN graphs can access and update, reducing overall memory overhead and management complexity.
Solution Approach 2:
The memory buffer acts as an intermediary between multiple TGN graphs. It mediates the interaction between different node types by providing a centralized storage and retrieval mechanism, allowing graphs to exchange information without direct complex memory management between them.
3Quantity of substance
If edge features are dynamically generated rather than pre-allocated, then memory efficiency is improved, but the computational complexity during execution increases
Solution Approach 1:
The patent implements dynamic edge feature generation where edge features are created on-demand during graph processing rather than being pre-allocated. This dynamic approach allows the system to adapt memory usage to actual computational needs, improving efficiency while the underlying computational complexity is managed through the structured TGN framework.
Data Source
AI summary
Discussed herein are devices, systems, and methods for more flexible temporal graph network (TGN) graph interaction. A method includes executing first and second temporal graph networks (TGNs) to generate embeddings of respective first and second dynamic graphs, storing, as respective edge features of a first node of the first graph and a second node of the second graph, a memory state vector of the first node and a memory state vector of the second node, and determining, based on the embeddings and the edge features, a likelihood of an edge between nodes of the first graph.