Polygonal Unadjusted Langevin Algorithms: Creating stable and efficient adaptive algorithms for neural networks

Dong-Young Lim, Sotirios Sabanis

Research output: Contribution to journalArticlepeer-review

Abstract

We present a new class of adaptive stochastic optimization algorithms, which overcomes many of the known shortcomings of popular adaptive optimizers that are currently used for the fine tuning of artificial neural networks (ANNs). Its underpinning theory relies on advances of Euler's polygonal approximations for stochastic differential equations (SDEs) with monotone coefficients. As a result, it inherits the stability properties of tamed algorithms, while it addresses other known issues, e.g. vanishing gradients in ANNs. In particular, we provide an nonasymptotic analysis and full theoretical guarantees for the convergence properties of an algorithm of this novel class, which we named TH$\varepsilon$O POULA (or, simply, TheoPouLa). Finally, several experiments are presented with different types of ANNs, which show the superior performance of TheoPouLa over many popular adaptive optimization algorithms.
Original languageEnglish
Pages (from-to)1-52
JournalJournal of Machine Learning Research
Volume25
Issue number53
Publication statusPublished - 28 Feb 2024

Keywords / Materials (for Non-textual outputs)

  • cs.LG
  • math.OC
  • math.PR
  • stat.ML

Fingerprint

Dive into the research topics of 'Polygonal Unadjusted Langevin Algorithms: Creating stable and efficient adaptive algorithms for neural networks'. Together they form a unique fingerprint.

Cite this