Hybrid Parallel Classifiers for Semantic Subspace Learning

Subspace learning is very important in today's world of information overload. Distinguishing between categories within a subset of a large data repository such as the web and the ability to do so in real time is critical for a successful search technique. The characteristics of data belonging to different domains are also varying widely. This merits the need for an architecture which caters to the differing characteristics of different data domains. In this paper we present a novel hybrid parallel architecture using different types of classifiers trained on different subspaces. We further compare the performance of our hybrid architecture with a single classifier – full data space learning and show that it outperforms the single classifier system by a large margin when tested with a variety of hybrid combinations. Our results show that subspace classification accuracy is boosted and learning time reduced significantly with this new hybrid architecture.