by M. MutnÃ½, M. Derezisnki, A. Krause

Abstract:

We analyze the convergence rate of the randomized Newton-like method introduced by Qu et. al. (2016) for smooth and convex objectives, which uses random coordinate blocks of a Hessian-over-approximation matrix $\bM$ instead of the true Hessian. The convergence analysis of the algorithm is challenging because of its complex dependence on the structure of $\bM$. However, we show that when the coordinate blocks are sampled with probability proportional to their determinant, the convergence rate depends solely on the eigenvalue distribution of matrix $\bM$, and has an analytically tractable form. To do so, we derive a fundamental new expectation formula for determinantal point processes. We show that determinantal sampling allows us to reason about the optimal subset size of blocks in terms of the spectrum of $\bM$. Additionally, we provide a numerical evaluation of our analysis, demonstrating cases where determinantal sampling is superior or on par with uniform sampling.

Reference:

Convergence Analysis of Block Coordinate Algorithms with Determinantal Sampling M. MutnÃ½, M. Derezisnki, A. KrauseIn Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS), AISTATS, 2020

Bibtex Entry:

@inproceedings{Mutny2020b, Author = {Mutn\'{y}, Mojmir and Derezisnki, Michal and Krause, Andreas}, Booktitle = {Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS)}, Month = {June}, Publisher = {AISTATS}, Title = {Convergence Analysis of Block Coordinate Algorithms with Determinantal Sampling}, Year = {2020}}