Graph neural network explanations made interactive and queryable for large data
Declarative Explanations for Graph Neural Networks: A Demonstration
Machine Learning
Summary
Graph Neural Networks (GNNs) help analyze complex network data but can be hard to understand because of their many layers. The authors created SliceGXQ, a tool that lets users interactively explore and explain how GNNs work on large graphs. People can use it to find mistakes, see which parts of the model influence decisions, and debug or analyze big network problems like fraud. The tool uses an easy language to ask questions and shows results visually.
What this means in practice
- •For data science teams: Investigate and debug graph neural network models by interactively examining layer explanations and node behaviors in large network datasets.
- •For fraud detection analysts: Identify suspicious nodes and relevant model features in large transactional graphs through targeted querying and visualization of neural network explanations.
Authors
Cibo Yu, Haitong Tang, Yinghui Wu, Arijit Khan, Tingting Zhu, Tingyang Chen, Xiangyu Ke
Abstract
Graph Neural Networks (GNNs) have proven effective for graph analytical tasks, yet their "black-box" nature complicates the understanding of their layer-wise behavior and hinders their application to large graphs. We demonstrate SliceGXQ, an end-to-end declarative and interactive system for querying and exploring layer-wise explanations for GNNs over large-scale graphs. In particular, SliceGXQ provides (1) an interactive graphical user interface for dataset, model, and explanation configuration; (2) a lightweight SPARQL-like parser that allows users to specify target nodes, layers, explanation modes, and constraints; and (3) an integrated execution and visualization workflow that invokes SliceGX as its underlying explanation engine. We demonstrate how users can diagnose a misclassified node, identify problematic layers and influential messages, and perform targeted model debugging, together with large-scale graph interpretation and fraud-detection analysis. Our demonstration video and source code are available at https://github.com/jinzhaoqet/SliceGXQ.