Locality Preserving Discriminative Dictionary Learning

The goal of dictionary learning is to learn an over-complete set of signals, called the dictionary, that can be used to represent signals of interest through sparse linear combinations. Recently, there has been growing interest in using dictionary learning for classification problems in computer vision. Representing natural signals successfully in a sparse manner, relies heavily on the dictionary used

In this paper, a novel discriminative dictionary learning approach is proposed that attempts to preserve the local structure of the data while encouraging discriminability. The reconstruction error and sparsity inducing $\ell_1$-penalty of dictionary learning are minimized alongside a locality preserving and discriminative term. In this setting, each data point is represented by a sparse linear combination of dictionary atoms with the goal that its $k$-nearest same-label neighbors are preserved. Since the class of a new data point is unknown, its sparse representation is found once for each class. The class that produces the lowest error is associated with that point. Experimental results show that this method outperforms state-of-the-art classifiers, especially when the training data is limited.

Related Publication:

  • S. Haghiri, H. R. Rabiee, A. Soltani-Farani, S. A. Hosseini, M. Shadloo, “Locality Preserving Discriminative Dictionary Learning”, IEEE Int. Conf. Image Processing, 2014

Project code:

  • The project code will be uploaded here soon.

People involved:

  • Siavash Haghiri, Hamid R. Rabiee, Ali Soltani-Farani, Seyyed Abbas Hosseini, Maryam Shadloo

Sparse Signal Processing Group