Augmented Hashing for Semi-Supervised Scenarios

Abstract. Hashing methods for fast approximate nearest-neighbor search are getting more and more attention with the excessive growth of the available data today. Embedding the points into the Hamming space is an important question of the hashing process. Analogously to machine learning there exist unsupervised, supervised and semi-supervised hash-ing methods. In this paper we propose a generic procedure to extend unsupervised codeword generators using error correcting codes and semi-supervised classifiers. To show the effectiveness of the method we combine linear spectral hashing and two semi-supervised algorithms in the experi-ments.