MEGAN: Multi-Explanation Graph Attention Network (2022-11-23T00:00:00.000000Z)

TL;DR

This work proposes a multi-explanation graph attention network (MEGAN) that can produce node and edge attributional explanations along multiple channels, the number of which is independent of task specifications, and finds that the model produces sparse high-fidelity explanations consistent with human intuition about those tasks.

Abstract

We propose a multi-explanation graph attention network (MEGAN). Unlike existing graph explainability methods, our network can produce node and edge attributional explanations along multiple channels, the number of which is independent of task specifications. This proves crucial to improve the interpretability of graph regression predictions, as explanations can be split into positive and negative evidence w.r.t to a reference value. Additionally, our attention-based network is fully differentiable and explanations can actively be trained in an explanation-supervised manner. We first validate our model on a synthetic graph regression dataset with known ground-truth explanations. Our network outperforms existing baseline explainability methods for the single- as well as the multi-explanation case, achieving near-perfect explanation accuracy during explanation supervision. Finally, we demonstrate our model's capabilities on multiple real-world datasets. We find that our model produces sparse high-fidelity explanations consistent with human intuition about those tasks.

Authors

Pascal Friederich

5 papers

Patrick Reiser

3 papers

Luca Torresi

2 papers

TL;DR

Abstract

Authors

References54 items

PAGE: Prototype-Based Model-Level Explanations for Graph Neural Networks

Encoding Concepts in Graph Neural Networks

BAGEL: A Benchmark for Assessing Graph Neural Network Explanations

GRETEL: Graph Counterfactual Explanation Evaluation Framework

DT+GNN: A Fully Explainable Graph Neural Network using Decision Trees

Learning to Scaffold: Optimizing Model Explanations for Teaching

Learning and Evaluating Graph Neural Network Explanations based on Counterfactual and Factual Reasoning

PLAM: A plug-in module for flexible graph attention learning

ProtGNN: Towards Self-Explaining Graph Neural Networks

CYBORG: Blending Human Saliency Into the Loss Improves Deep Learning-Based Synthetic Face Detection

GNES: Learning to Explain Graph Neural Networks

Towards Self-Explainable Graph Neural Network

GCExplainer: Human-in-the-Loop Concept-based Explanations for Graph Neural Networks

How Attentive are Graph Attention Networks?

Zorro: Valid, Sparse, and Stable Explanations in Graph Neural Networks

Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity

Supervising Model Attention with Human Explanations for Robust Natural Language Inference

Explainability in Graph Neural Networks: A Taxonomic Survey

Evaluating Explanations: How Much Do Explanations from the Teacher Aid Students?

Parameterized Explainer for Graph Neural Network

Scientific intuition inspired by machine learning-generated hypotheses

Pushing the limits of solubility prediction via quality-oriented data selection

Drug discovery with explainable artificial intelligence

GraphLIME: Local Interpretable Model Explanations for Graph Neural Networks

ERASER: A Benchmark to Evaluate Rationalized NLP Models

Layerwise Relevance Visualization in Convolutional Text Graph Classifiers

Learning to Deceive with Attention-Based Explanations

AqSolDB, a curated reference set of aqueous solubility and 2D descriptors for a diverse set of compounds

Explainability Methods for Graph Convolutional Neural Networks

Explainable Machine Learning for Scientific Insights and Discoveries

GNNExplainer: Generating Explanations for Graph Neural Networks

Reliable and explainable machine-learning methods for accelerated material discovery

Using attribution to decode binding mechanism in neural network models for chemistry

Sanity Checks for Saliency Maps

Learning what and where to attend

The (Un)reliability of saliency methods

Graph Attention Networks

Exploring Human-like Attention Supervision in Visual Question Answering

MoleculeNet: a benchmark for molecular machine learning

Towards A Rigorous Science of Interpretable Machine Learning

Design of efficient molecular organic light-emitting diodes by a high-throughput virtual screening and experimental approach.

Semi-Supervised Classification with Graph Convolutional Networks

GloVe: Global Vectors for Word Representation

Design of efficient thermally activated delayed fluorescence materials for pure blue organic light emitting diodes.

Efficient up-conversion of triplet excitons into a singlet state and its application for organic light emitting diodes

Estimation of Aqueous Solubility for a Diverse Set of Organic Compounds Based on Molecular Topology.

ESOL: Estimating Aqueous Solubility Directly from Molecular Structure

trialyamine bridges between the fragments

2021) already investigates this property using an interpretable decision tree approach

The MEGAN models used in this first experiment use only the one explanation channel

Graph neural networks in TensorFlowKeras with RaggedTensor representation (kgcnn)

Evaluating Attribution for Graph Neural Networks

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

RDKit: Open-source cheminformatics

Field of Study