Discriminative Bayesian Dictionary Learning for Classification
This work addresses the challenge of improving classification accuracy in computer vision tasks, such as face and action recognition, by developing a novel Bayesian method for discriminative dictionary learning, which is incremental as it builds upon existing sparse representation approaches.
The authors tackled the problem of learning discriminative dictionaries for classification by proposing a Bayesian approach that infers probability distributions over dictionary atoms using a Beta Process and associates class labels via Bernoulli distributions, resulting in consistent outperformance over state-of-the-art methods in experiments on face and action recognition and object and scene-category classification across five public datasets.
We propose a Bayesian approach to learn discriminative dictionaries for sparse representation of data. The proposed approach infers probability distributions over the atoms of a discriminative dictionary using a Beta Process. It also computes sets of Bernoulli distributions that associate class labels to the learned dictionary atoms. This association signifies the selection probabilities of the dictionary atoms in the expansion of class-specific data. Furthermore, the non-parametric character of the proposed approach allows it to infer the correct size of the dictionary. We exploit the aforementioned Bernoulli distributions in separately learning a linear classifier. The classifier uses the same hierarchical Bayesian model as the dictionary, which we present along the analytical inference solution for Gibbs sampling. For classification, a test instance is first sparsely encoded over the learned dictionary and the codes are fed to the classifier. We performed experiments for face and action recognition; and object and scene-category classification using five public datasets and compared the results with state-of-the-art discriminative sparse representation approaches. Experiments show that the proposed Bayesian approach consistently outperforms the existing approaches.