Robust Classification for (Dis)Similarity Matrices

• Activity Number: 343
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Cencheng Shen*+ and Li Chen and Carey Priebe
• Companies: Johns Hopkins University and Johns Hopkins University and Johns Hopkins University
• Keywords: robust classifier ; sparse representation ; similarity matrix ; stochastic block model ; consistency

Abstract

We develop a robust classification methodology for data given by (dis)similarity matrices. The algorithm finds a sparse representation of testing data in terms of training data, using the (dis)similarity observations directly without the need for embedding. We demonstrate the efficacy of our approach using observed (dis)similarity data such as text data, video data, and graph data. And we investigate the consistency and robustness of our classifier theoretically for dot product similarity matrices and stochastic block models adjacency matrices.