22
1
v1v2 (latest)

Dictionary Learning for the Almost-Linear Sparsity Regime

Abstract

Dictionary learning, the problem of recovering a sparsely used matrix DRM×K\mathbf{D} \in \mathbb{R}^{M \times K} and NN ss-sparse vectors xiRK\mathbf{x}_i \in \mathbb{R}^{K} from samples of the form yi=Dxi\mathbf{y}_i = \mathbf{D}\mathbf{x}_i, is of increasing importance to applications in signal processing and data science. When the dictionary is known, recovery of xi\mathbf{x}_i is possible even for sparsity linear in dimension MM, yet to date, the only algorithms which provably succeed in the linear sparsity regime are Riemannian trust-region methods, which are limited to orthogonal dictionaries, and methods based on the sum-of-squares hierarchy, which requires super-polynomial time in order to obtain an error which decays in MM. In this work, we introduce SPORADIC (SPectral ORAcle DICtionary Learning), an efficient spectral method on family of reweighted covariance matrices. We prove that in high enough dimensions, SPORADIC can recover overcomplete (K>MK > M) dictionaries satisfying the well-known restricted isometry property (RIP) even when sparsity is linear in dimension up to logarithmic factors. Moreover, these accuracy guarantees have an ``oracle property" that the support and signs of the unknown sparse vectors xi\mathbf{x}_i can be recovered exactly with high probability, allowing for arbitrarily close estimation of D\mathbf{D} with enough samples in polynomial time. To the author's knowledge, SPORADIC is the first polynomial-time algorithm which provably enjoys such convergence guarantees for overcomplete RIP matrices in the near-linear sparsity regime.

View on arXiv
Comments on this paper