Using domain-specific knowledge in SVMs

Abstract

In this study we develop a methodology that incorporates domain-specific knowledge in Support Vector Machines (SVMs) to enhance generalization error bounds. We begin with a detailed literature review on encoding prior knowledge with SVMs. First, we consider prior knowledge about the domain by incorporating upper and lower bounds of attributes. We then consider a more general framework that allows us to encode prior knowledge in the form of linear constraints formed by attributes, and propose a slightly improved version of ellipsoid method to obtain tighter error bounds. Finally, by comprehensive comparative numerical analysis we compare the effectiveness of incorporating domain knowledge versus using SVMs error bounds obtained without incorporating domain knowledge.

Available Versions of this Item

• Using domain-specific knowledge in SVMs. (deposited 20 Dec 2006 02:00) [Currently Displayed]
  • Using Domain-Specific Knowledge in SVMs Generalization Error Bounds. (deposited 31 Oct 2007 09:25)