Deep Kernel Supervised Hashing for Node Classification in Structural Networks
This work solves the node classification problem for applications in structural networks, representing an incremental improvement over existing network embedding methods.
The paper tackles the problem of node classification in structural networks by addressing the linearly inseparable issue in low-dimensional space and the inability to incorporate both network structure and node label information simultaneously, resulting in a method that significantly outperforms state-of-the-art baselines on three real-world benchmark datasets.
Node classification in structural networks has been proven to be useful in many real world applications. With the development of network embedding, the performance of node classification has been greatly improved. However, nearly all the existing network embedding based methods are hard to capture the actual category features of a node because of the linearly inseparable problem in low-dimensional space; meanwhile they cannot incorporate simultaneously network structure information and node label information into network embedding. To address the above problems, in this paper, we propose a novel Deep Kernel Supervised Hashing (DKSH) method to learn the hashing representations of nodes for node classification. Specifically, a deep multiple kernel learning is first proposed to map nodes into suitable Hilbert space to deal with linearly inseparable problem. Then, instead of only considering structural similarity between two nodes, a novel similarity matrix is designed to merge both network structure information and node label information. Supervised by the similarity matrix, the learned hashing representations of nodes simultaneously preserve the two kinds of information well from the learned Hilbert space. Extensive experiments show that the proposed method significantly outperforms the state-of-the-art baselines over three real world benchmark datasets.