Analysis of the Optimization Landscapes for Overcomplete Representation Learning

Abstract

We study nonconvex optimization landscapes for learning overcomplete representations, including learning (i) sparsely used overcomplete dictionaries and (ii) convolutional dictionaries, where these unsupervised learning problems find many applications in high-dimensional data analysis. Despite the empirical success of simple nonconvex algorithms, theoretical justifications of why these methods work so well are far from satisfactory. In this work, we show these problems can be formulated as ${\ell }^4$-norm optimization problems with spherical constraint, and study the geometric properties of their nonconvex optimization landscapes. For both problems, we show the nonconvex objectives have benign (global) geometric structures, in the sense that every local minimizer is close to one of the target solutions and every saddle point exhibits negative curvature. This discovery enables the development of guaranteed global optimization methods using simple initializations. For both problems, we show the nonconvex objectives have benign geometric structures -- every local minimizer is close to one of the target solutions and every saddle point exhibits negative curvature -- either in the entire space or within a sufficiently large region. This discovery ensures local search algorithms (such as Riemannian gradient descent) with simple initializations approximately find the target solutions. Finally, numerical experiments justify our theoretical discoveries.

68 pages, 5 figures

References in corpus (11)

• Generalized power method for sparse principal component analysis
• Global rates of convergence for nonconvex optimization on manifolds
• Complete Dictionary Recovery over the Sphere I: Overview and the Geometric Picture
• Implicit Regularization in Nonconvex Statistical Estimation: Gradient Descent Converges Linearly for Phase Retrieval, Matrix Completion, and Blind Deconvolution
• Convolutional Dictionary Learning: A Comparative Review and New Algorithms
• Complete Dictionary Recovery over the Sphere II: Recovery by Riemannian Trust-region Method
• When Are Nonconvex Problems Not Scary?
• Complete Dictionary Recovery over the Sphere
• Analyzing Tensor Power Method Dynamics in Overcomplete Regime
• When Does Non-Orthogonal Tensor Decomposition Have No Spurious Local Minima?
• Short-and-Sparse Deconvolution -- A Geometric Approach

Cited by in corpus (2)

• From Symmetry to Geometry: Tractable Nonconvex Problems
• An Unconstrained Layer-Peeled Perspective on Neural Collapse