Robust training on approximated minimal-entropy set
This addresses the problem of sensor failure corrupting training data for practitioners in fields like multimodal sensing, though it appears incremental as it builds on existing robust classification frameworks.
The paper tackles robust binary classification in the presence of corrupted training data (anomalies) by proposing a joint method for classification and anomaly detection, resulting in improved classification accuracy and anomaly detection rates over previous methods.
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.