Extensions to Self-Taught Hashing: Kernelisation and Supervision

The ability of fast similarity search at large scale is of great importance to many Information Retrieval (IR) applications. A promising way to accelerate similarity search is semantic hashing which designs compact binary codes for a large number of documents so that semantically similar documents are mapped to similar codes (within a short Hamming distance). Since each bit in the binary code for a document can be regarded as a binary feature of it, semantic hashing is essentially a process of generating a few most informative binary features to represent the documents. Recently, we have proposed a novel Self-Taught Hashing (STH) approach to semantic hashing (that is going to be published in SIGIR-2010): we first find the optimal l-bit binary codes for all documents in the given corpus via unsupervised learning, and then train l classifiers via supervised learning to predict the l-bit code for any query document unseen before. In this paper, we present two further extensions to our STH technique: one is kernelisation (i.e., employing nonlinear kernels to achieve nonlinear hashing), and the other is supervision (i.e., exploiting the category label information to enhance the effectiveness of hashing). The advantages of these extensions have been shown through experiments on synthetic datasets and real-world datasets respectively.