Fast, Provably convergent IRLS Algorithm for p-norm Linear Regression

Abstract

Linear regression in ${\ell }_p$-norm is a canonical optimization problem that arises in several applications, including sparse recovery, semi-supervised learning, and signal processing. Generic convex optimization algorithms for solving ${\ell }_p$-regression are slow in practice. Iteratively Reweighted Least Squares (IRLS) is an easy to implement family of algorithms for solving these problems that has been studied for over 50 years. However, these algorithms often diverge for p > 3, and since the work of Osborne (1985), it has been an open problem whether there is an IRLS algorithm that is guaranteed to converge rapidly for p > 3. We propose p-IRLS, the first IRLS algorithm that provably converges geometrically for any $p\in [2,\infty ).$ Our algorithm is simple to implement and is guaranteed to find a $(1+\epsilon )$-approximate solution in $O(p^{3.5}{m}^{\frac{p-2}{2(p-1)}}\log \frac{m}{\epsilon })\le O_p(\sqrt{m}\log \frac{m}{\epsilon })$ iterations. Our experiments demonstrate that it performs even better than our theoretical bounds, beats the standard Matlab/CVX implementation for solving these problems by 10--50x, and is the fastest among available implementations in the high-accuracy regime.

Code for this work is available at https://github.com/utoronto-theory/pIRLS

References in corpus (8)

• Asymptotic behavior of ${\ell }_p$-based Laplacian regularization in semi-supervised learning
• Hamiltonian Descent Methods
• Algorithms for Lipschitz Learning on Graphs
• IRLS and Slime Mold: Equivalence and Convergence
• Fast minimization of structured convex quartics
• Iterative Design of L_p Digital Filters
• Nonlocal $p$-Laplacian Variational problems on graphs
• Faster p-norm minimizing flows, via smoothed q-norm problems

Cited by in corpus (3)

• Iteratively Reweighted Least Squares for Basis Pursuit with Global Linear Convergence Rate
• A General Approach for Computing a Consensus in Group Decision Making That Integrates Multiple Ethical Principles
• Acceleration with a Ball Optimization Oracle