Incremental Metapath Storage for Dynamic Graph Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing metapath-based heterogeneous graph neural networks face significant time delays and inefficiencies in real-time inference due to the time-consuming process of generating metapath instances, especially in dynamic heterogeneous graphs with frequent updates.
Innovation Solution
A method and system for incremental, dynamic maintenance of metapath instances, involving reformatting metapath instances into path graphs, performing graph updates, and applying merge and restoration operations based on triggering conditions to accelerate metapath matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pre-processing traversal is conducted for all vertices to match metapath instances, then complete metapath matching is achieved, but time delay increases significantly
Solution Approach 1:
The patent pre-calculates and stores metapath instance matching results in a cache structure before real-time inference is needed. When graph updates occur, only incremental updates to the cached results are performed rather than complete re-matching, significantly reducing real-time inference delay while maintaining matching completeness.
Solution Approach 2:
The patent divides the graph into multiple partitions and processes metapath matching in parallel across different partitions. This segmentation allows simultaneous processing of multiple vertex subsets, reducing overall matching time while ensuring all metapath instances are eventually matched.
2Productivity
If candidate set modification is used based on existing metapath instances, then processing speed improves, but model accuracy decreases when graph updates increase
Solution Approach 1:
The patent implements a feedback mechanism where the system monitors graph update frequency and magnitude. When updates exceed a threshold, the system automatically triggers a full re-matching process to refresh the candidate set, ensuring model accuracy is maintained while normally operating in the faster incremental update mode.
Solution Approach 2:
The patent dynamically adjusts the metapath instance generation strategy based on current graph state. For small updates, it uses efficient candidate set modification; for large updates, it transitions to complete re-matching. This dynamic adaptation maintains both speed and accuracy under different operating conditions.
3Reliability
If all metapath instances are used for inference, then model accuracy is maximized, but computational complexity increases
Solution Approach 1:
The patent assigns different qualities or priorities to different metapath instances based on their relevance to specific inference tasks. Instead of uniformly processing all metapath instances, the system identifies and prioritizes the most important ones for each particular inference scenario, reducing computational complexity while maintaining accuracy for critical metapaths.
Data Source
AI summary
A method for incremental metapath storage and dynamic maintenance is provided, which includes, reformatting metapath instances, from a designated heterogeneous graph and of a designated metapath type, into path graphs; executing graph updating tasks and performing dynamic maintenance on the updated path graphs, traversing the path graph to obtain the location of metapath updates and update the path graph; for metapaths with length greater than 2 and with symmetrical central portion, central merge operation is performed to simplify path graph and perform subsequent restoration operation; and directly perform restoration operation on path graphs that do not meet the merging conditions. The present disclosure utilizes characteristics of graph update to obtain locality of metapath updates, and combines internal relationship characteristics of metapath instances to greatly speed up metapath generation and achieve real-time inference of dynamic heterogeneous graph models.