Class-wise deep dictionary learning

Abstract

In this work we propose a classification framework called class-wise deep dictionary learning (CWDDL). For each class, multiple levels of dictionaries are learnt using features from the previous level as inputs (for first level the input is the raw training sample). It is assumed that the cascaded dictionaries form a basis for expressing test samples for that class in the sparsest form. Based on this assumption sparse representation based classification (SRC) is employed. Experiments have been carried out on some benchmark deep learning datasets (MNIST and its variations and energy data); our proposed method has been compared with Deep Belief Network, Stacked Auto encoder and Label Consistent KSVD (dictionary learning). We find that our proposed method yields better results than these techniques and requires much smaller run-times. We have compared our method with the best-in-class results for these datasets; we find that our class-wise deep dictionary learning (CWDDL) approach features in the top 10 results.