DOC3-Deep One Class Classification using Contradictions

TL;DR

This paper proposes DOC3, a deep one-class classification method that leverages contradictions (Universum learning) to improve generalization, validated through theoretical analysis and empirical experiments on real datasets.

Contribution

It formalizes the use of contradictions in deep one-class classification and introduces the DOC3 algorithm, demonstrating lower generalization error.

Findings

DOC3 outperforms baseline algorithms on real datasets.

Learning from contradictions reduces generalization error.

Theoretical analysis shows lower Empirical Rademacher Complexity for DOC3.

Abstract

This paper introduces the notion of learning from contradictions (a.k.a Universum learning) for deep one class classification problems. We formalize this notion for the widely adopted one class large-margin loss, and propose the Deep One Class Classification using Contradictions (DOC3) algorithm. We show that learning from contradictions incurs lower generalization error by comparing the Empirical Rademacher Complexity (ERC) of DOC3 against its traditional inductive learning counterpart. Our empirical results demonstrate the efficacy of DOC3 compared to popular baseline algorithms on several real-life data sets.