An agent model for information filtering using revolutionary RSVD technique

Abstract

© 2014, Chiang Mai University. All rights reserved. This paper proposes a collaborative software agent model. The agent works in a distributed environment making recommendation based on its up-to-date knowledge. This knowledge is partly acquired from other collaborative agents to combine with its own prior knowledge by means of a revolutionary regularized singular value decomposition (rRSVD) technique. The technique is used as an adaptation process for the agent to learn and update the knowledge periodically. This process employs one of the three agent adaptation models, namely, 2-phase, 1-phase, or non-adaptation that is suitable for the operating bandwidth, along with a fast incremental knowledge adaptation algorithm. As a consequence, the adapted agent will be able to work alone in a distributed environment at a satisfactory level of performance.