Learning to classify with possible sensor failures

TL;DR

This paper introduces GEM-MED, a robust classification framework that jointly learns to classify and detect anomalies caused by sensor failures, improving accuracy and detection rates in contaminated datasets.

Contribution

The paper presents a novel GEM-MED method combining large-margin classification with anomaly detection using a non-parametric regularizer, enhancing robustness against sensor failures.

Findings

GEM-MED outperforms previous methods in accuracy and detection rate.

Effective in both simulated and real multimodal datasets.

Joint classification and anomaly detection improves overall robustness.

Abstract

In this paper, we propose a general framework to learn a robust large-margin binary classifier when corrupt measurements, called anomalies, caused by sensor failure might be present in the training set. The goal is to minimize the generalization error of the classifier on non-corrupted measurements while controlling the false alarm rate associated with anomalous samples. By incorporating a non-parametric regularizer based on an empirical entropy estimator, we propose a Geometric-Entropy-Minimization regularized Maximum Entropy Discrimination (GEM-MED) method to learn to classify and detect anomalies in a joint manner. We demonstrate using simulated data and a real multimodal data set. Our GEM-MED method can yield improved performance over previous robust classification methods in terms of both classification accuracy and anomaly detection rate.