The paper introduces DIFFPOOL, a differentiable graph pooling module designed to generate hierarchical representations of graphs, addressing the limitation of current graph neural networks (GNNs) that are inherently flat and lack hierarchical structure. DIFFPOOL learns a soft cluster assignment for nodes at each layer of a deep GNN, mapping nodes to clusters that form the input for the next GNN layer. This process is repeated for multiple layers, resulting in a hierarchical representation of the graph. The authors demonstrate that combining existing GNN methods with DIFFPOOL improves accuracy by 5-10% on graph classification benchmarks, achieving state-of-the-art results on four out of five datasets. The method is shown to learn interpretable hierarchical clusters that correspond to well-defined communities in the input graphs. The paper also discusses related work, experimental results, and future directions, including the potential for learning hard cluster assignments and applying hierarchical pooling to other tasks.The paper introduces DIFFPOOL, a differentiable graph pooling module designed to generate hierarchical representations of graphs, addressing the limitation of current graph neural networks (GNNs) that are inherently flat and lack hierarchical structure. DIFFPOOL learns a soft cluster assignment for nodes at each layer of a deep GNN, mapping nodes to clusters that form the input for the next GNN layer. This process is repeated for multiple layers, resulting in a hierarchical representation of the graph. The authors demonstrate that combining existing GNN methods with DIFFPOOL improves accuracy by 5-10% on graph classification benchmarks, achieving state-of-the-art results on four out of five datasets. The method is shown to learn interpretable hierarchical clusters that correspond to well-defined communities in the input graphs. The paper also discusses related work, experimental results, and future directions, including the potential for learning hard cluster assignments and applying hierarchical pooling to other tasks.