Learning Deep Representations for Graph Clustering

• DOI: 10.1609/aaai.v28i1.8916
• Publication Date: 2022-06-23T09:51:49Z
• Authors: Fei Tian; Bin Gao; Qing Cui; Enhong Chen; Tie-Yan Liu
• Keywords: Autoencoder; Spectral Clustering; Graph Embedding

Recently deep learning has been successfully adopted in many applications such as speech recognition and image classification. In this work, we explore the possibility of employing graph clustering. We propose a simple method, which first learns nonlinear embedding original by stacked autoencoder, then runs $k$-means algorithm on to obtain clustering result. show that method solid theoretical foundation, due similarity between autoencoder spectral terms what they actually optimize. Then, demonstrate proposed is more efficient flexible than First, computational complexity much lower clustering: former can be linear number nodes sparse while latter super quadratic eigenvalue decomposition. Second, when additional sparsity constraint imposed, simply employ developed literature learning; however, it non-straightforward implement method. The experimental results various datasets significantly outperforms conventional clearly indicates effectiveness