In this paper, we consider multi-sensor classification when there is a largenumber of unlabeled samples. The problem is formulated under the multi-viewlearning framework and a Consensus-based Multi-View Maximum EntropyDiscrimination (CMV-MED) algorithm is proposed. By iteratively maximizing thestochastic agreement between multiple classifiers on the unlabeled dataset, thealgorithm simultaneously learns multiple high accuracy classifiers. Wedemonstrate that our proposed method can yield improved performance overprevious multi-view learning approaches by comparing performance on three realmulti-sensor data sets.
展开▼
