TATi-Thermodynamic Analytics ToolkIt: TensorFlow-based software for posterior sampling in machine learning applications

Frederik Heber, Zofia Trstanova, Benedict Leimkuhler

Research output: Contribution to journalArticlepeer-review

Abstract

With the advent of GPU-assisted hardware and maturing high-efficiency software platformssuch as TensorFlow and PyTorch, Bayesian posterior sampling for neural networks becomes plausible. In this article we discuss Bayesian parametrization in machine learning based on Markov Chain Monte Carlo methods, specifically discretized stochastic differential equations such as Langevin dynamics and extended system methods in which an ensemble of walkers is employed to enhance sampling. We provide a glimpse of the potential of the sampling-intensive approach by studying (and visualizing) the loss landscape of a neural network applied to the MNIST data set. Moreover, we investigate how the sampling efficiency itself can be significantly enhanced through an ensemble quasi-Newton preconditioning method. This article accompanies the release of a new TensorFlow software package, the ThermodynamicAnalytics ToolkIt, which is used in the computational experiment
Original languageEnglish
Number of pages25
JournalJournal of Machine Learning Research
Publication statusAccepted/In press - 14 Jul 2020

Fingerprint Dive into the research topics of 'TATi-Thermodynamic Analytics ToolkIt: TensorFlow-based software for posterior sampling in machine learning applications'. Together they form a unique fingerprint.

Cite this