Timezone: »

Low-Precision Stochastic Gradient Langevin Dynamics
Ruqi Zhang · Andrew Wilson · Christopher De Sa

Tue Jul 19 02:30 PM -- 02:35 PM (PDT) @ Room 307

While low-precision optimization has been widely used to accelerate deep learning, low-precision sampling remains largely unexplored. As a consequence, sampling is simply infeasible in many large-scale scenarios, despite providing remarkable benefits to generalization and uncertainty estimation for neural networks. In this paper, we provide the first study of low-precision Stochastic Gradient Langevin Dynamics (SGLD), showing that its costs can be significantly reduced without sacrificing performance, due to its intrinsic ability to handle system noise. We prove that the convergence of low-precision SGLD with full-precision gradient accumulators is less affected by the quantization error than its SGD counterpart in the strongly convex setting. To further enable low-precision gradient accumulators, we develop a new quantization function for SGLD that preserves the variance in each update step. We demonstrate that low-precision SGLD achieves comparable performance to full-precision SGLD with only 8 bits on a variety of deep learning tasks.

Author Information

Ruqi Zhang (UT Austin/Purdue)
Andrew Wilson (New York University)
Andrew Wilson

Andrew Gordon Wilson is faculty in the Courant Institute and Center for Data Science at NYU. His interests include probabilistic modelling, Gaussian processes, Bayesian statistics, physics inspired machine learning, and loss surfaces and generalization in deep learning. His webpage is https://cims.nyu.edu/~andrewgw.

Christopher De Sa (Cornell)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors