How Much Restricted Isometry is Needed In Nonconvex Matrix Recovery?

Abstract

When the linear measurements of an instance of low-rank matrix recovery satisfy a restricted isometry property (RIP)---i.e. they are approximately norm-preserving---the problem is known to contain no spurious local minima, so exact recovery is guaranteed. In this paper, we show that moderate RIP is not enough to eliminate spurious local minima, so existing results can only hold for near-perfect RIP. In fact, counterexamples are ubiquitous: we prove that every x is the spurious local minimum of a rank-1 instance of matrix recovery that satisfies RIP. One specific counterexample has RIP constant $\delta =1/2$, but causes randomly initialized stochastic gradient descent (SGD) to fail 12% of the time. SGD is frequently able to avoid and escape spurious local minima, but this empirical result shows that it can occasionally be defeated by their existence. Hence, while exact recovery guarantees will likely require a proof of no spurious local minima, arguments based solely on norm preservation will only be applicable to a narrow set of nearly-isotropic instances.

arXiv comment32nd Conference on Neural Information Processing Systems (NIPS 2018)

References in corpus (8)

• Guaranteed Minimum-Rank Solutions of Linear Matrix Equations via Nuclear Norm Minimization
• Understanding deep learning requires rethinking generalization
• The Marginal Value of Adaptive Gradient Methods in Machine Learning
• A Reliable Effective Terascale Linear Learning System
• Consistency of trace norm minimization
• Matrix Completion has No Spurious Local Minimum
• Global Optimality in Low-rank Matrix Optimization
• Gradient Descent Can Take Exponential Time to Escape Saddle Points

Cited by in corpus (1)

• Inference and Uncertainty Quantification for Noisy Matrix Completion