On the Sample Complexity of Predictive Sparse Coding
Abstract
The goal of predictive sparse coding is to learn a representation of examples as sparse linear combinations of elements from a dictionary, such that a learned hypothesis linear in the new representation performs well on a predictive task. Predictive sparse coding algorithms recently have demonstrated impressive performance on a variety of supervised tasks, but their generalization properties have not been studied. We establish the first generalization error bounds for predictive sparse coding, covering two settings: 1) the overcomplete setting, where the number of features k exceeds the original dimensionality d; and 2) the high or infinite-dimensional setting, where only dimension-free bounds are useful. Both learning bounds intimately depend on stability properties of the learned sparse encoder, as measured on the training sample. Consequently, we first present a fundamental stability result for the LASSO, a result characterizing the stability of the sparse codes with respect to perturbations to the dictionary. In the overcomplete setting, we present an estimation error bound that decays as \tilde{O}(sqrt(d k/m)) with respect to d and k. In the high or infinite-dimensional setting, we show a dimension-free bound that is \tilde{O}(sqrt(k^2 s / m)) with respect to k and s, where s is an upper bound on the number of non-zeros in the sparse code for any training data point.
Cite
@article{arxiv.1202.4050,
title = {On the Sample Complexity of Predictive Sparse Coding},
author = {Nishant A. Mehta and Alexander G. Gray},
journal= {arXiv preprint arXiv:1202.4050},
year = {2012}
}
Comments
Sparse Coding Stability Theorem from version 1 has been relaxed considerably using a new notion of coding margin. Old Sparse Coding Stability Theorem still in new version, now as Theorem 2. Presentation of all proofs simplified/improved considerably. Paper reorganized. Empirical analysis showing new coding margin is non-trivial on real datasets