Latent linkage semantic kernels for collective classification of link data

摘要: Generally, links among objects demonstrate certain patterns and contain rich semantic clues. These important clues can be used to improve classification accuracy. However, many real-world link data may exhibit more complex regularity. For example, there some noisy that carry no human editorial endorsement about relationships. To effectively capture such regularity, this paper proposes latent linkage kernels (LLSKs) by first introducing the model local global dependency structure of a graph then applying singular value decomposition (SVD) in kernel-induced space. computational efficiency on large datasets, we also develop block-based algorithm for LLSKs. A kernel-based contextual network (KCDN) is presented exploit dependencies collective classification. We provide experimental results demonstrating KCDN model, together with LLSKs, demonstrates relatively high robustness datasets computation method scale well varying sizes problem.