Subgraph Ranking for Graph Classification Explainability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models, particularly graph classification models, lack explainability, making it difficult to trust and validate their decisions, especially in critical applications like medicine and recommendation systems where understanding the reasoning behind predictions is essential.
Innovation Solution
A method and system for ranking subgraphs as potential explanations for graph classification by merging and embedding graphs using a machine learning model, calculating expressivity scores, and ranking subgraphs to provide interpretable explanations for labeled edge types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If graph classification models are used for accurate predictions, then prediction accuracy is improved, but explainability of decisions deteriorates
Solution Approach 1:
The patent segments the graph into multiple subgraphs and ranks them by their contribution to the classification decision. This allows the system to identify and present only the relevant subgraphs that contribute to the prediction, providing explainability without sacrificing the overall prediction accuracy of the graph classification model.
Solution Approach 2:
The patent introduces an intermediary explanation system that acts as a mediator between the black-box graph classification model and the user. This intermediary generates subgraph explanations that bridge the gap between the model's internal decision-making process and human understanding, maintaining prediction accuracy while improving explainability.
2Ease of operation
If detailed explanations are provided for model decisions, then interpretability is improved, but computational complexity deteriorates
Solution Approach 1:
The patent extracts only the essential subgraphs that contribute to the classification decision, rather than providing explanations for the entire graph. This extraction approach reduces computational complexity by focusing only on the relevant portions of the graph while still providing sufficient interpretability for understanding the model's decision.
Solution Approach 2:
The patent applies local quality by providing detailed explanations only for the specific subgraphs that are relevant to the classification decision, rather than uniformly explaining the entire graph. This localized approach to explanation reduces overall computational complexity while maintaining interpretability where it matters most.
3Loss of information
If subgraph ranking is performed to provide explanations, then explainability is improved, but processing time deteriorates
Solution Approach 1:
The patent performs partial action by ranking only the top-k most relevant subgraphs rather than analyzing and ranking all possible subgraphs. This partial approach provides sufficient explainability for understanding the model's decision while significantly reducing the processing time required for subgraph ranking.
Data Source
AI summary
There is provided a method and a system for ranking subgraphs as potential explanations for a labelled edge type class. A first graph representing labelled digital items is obtained, where each labelled digital item is represented as an entity node connected via a labelled edge type to a property value node. The first graph is combined with a second graph representing structured relations in the labelled digital items to obtain a combined graph. Unlabelled digital items are received and matched to respective subgraphs in the combined graph. A machine learning model is used to embed the combined graph to generate graph vectors, and an expressivity score between matched subgraphs and respective labelled edge types based on the generated graph vectors. The matched subgraphs are ranked based on the expressivity score to obtain a ranked set of subgraphs as potential explanations for a respective labelled edge type class.