Maximum Entropy Weighted Independent Set Pooling for Graph Neural Networks

In this paper, we propose a novel pooling layer for graph neural networks based on maximizing the mutual information between the pooled graph and the input graph. Since the maximum mutual information is difficult to compute, we employ the Shannon capacity of a graph as an inductive bias to our pooling method. More precisely, we show that the input graph to the pooling layer can be viewed as a representation of a noisy communication channel. For such a channel, sending the symbols belonging to an independent set of the graph yields a reliable and error-free transmission of information. We show that reaching the maximum mutual information is equivalent to finding a maximum weight independent set of the graph where the weights convey entropy contents. Through this communication theoretic standpoint, we provide a distinct perspective for posing the problem of graph pooling as maximizing the information transmission rate across a noisy communication channel, implemented by a graph neural network. We evaluate our method, referred to as Maximum Entropy Weighted Independent Set Pooling (MEWISPool), on graph classification tasks and the combinatorial optimization problem of the maximum independent set. Empirical results demonstrate that our method achieves the state-of-the-art and competitive results on graph classification tasks and the maximum independent set problem in several benchmark datasets.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Graph Classification COLLAB MEWISPool Accuracy 79.66% # 15
Graph Classification D&D MEWISPool Accuracy 84.33% # 2
Graph Classification FRANKENSTEIN MEWISPool Accuracy 73.46 # 2
Graph Classification IMDb-B MEWISPool Accuracy 82.13% # 2
Graph Classification IMDb-M MEWISPool Accuracy 56.23% # 4
Graph Classification MUTAG MEWISPool Accuracy 96.66% # 2
Graph Classification Mutagenicity MEWISPool Accuracy 80.66 # 3
Graph Classification PROTEINS MEWISPool Accuracy 80.71% # 5

Methods


No methods listed for this paper. Add relevant methods here