Adaptive Importance Sampling for Finite-Sum Optimization and Sampling with Decreasing Step-Sizes
Hanchi, Ayoub El, Stephens, David A.
Reducing the variance of the gradient estimator is known to improve the convergence rate of stochastic gradient-based optimization and sampling algorithms. One way of achieving variance reduction is to design importance sampling strategies. Recently, the problem of designing such schemes was formulated as an online learning problem with bandit feedback, and algorithms with sub-linear static regret were designed. In this work, we build on this framework and propose Avare, a simple and efficient algorithm for adaptive importance sampling for finite-sum optimization and sampling with decreasing step-sizes. Under standard technical conditions, we show that Avare achieves $\mathcal{O}(T^{2/3})$ and $\mathcal{O}(T^{5/6})$ dynamic regret for SGD and SGLD respectively when run with $\mathcal{O}(1/t)$ step sizes. We achieve this dynamic regret bound by leveraging our knowledge of the dynamics defined by the algorithm, and combining ideas from online learning and variance-reduced stochastic optimization. We validate empirically the performance of our algorithm and identify settings in which it leads to significant improvements.
Mar-22-2021
- Country:
- Oceania > Australia
- New South Wales > Sydney (0.04)
- North America
- United States
- District of Columbia > Washington (0.04)
- Washington > King County
- Bellevue (0.04)
- New York
- Richmond County > New York City (0.04)
- Queens County > New York City (0.04)
- New York County > New York City (0.04)
- Kings County > New York City (0.04)
- Bronx County > New York City (0.04)
- Nevada > Clark County
- Las Vegas (0.04)
- California
- San Francisco County > San Francisco (0.14)
- Los Angeles County > Long Beach (0.14)
- Canada
- Quebec > Montreal (0.14)
- British Columbia > Metro Vancouver Regional District
- Vancouver (0.04)
- United States
- Europe
- Oceania > Australia
- Genre:
- Research Report (1.00)
- Industry:
- Education (0.74)
- Technology: