Interacting Contour Stochastic Gradient Langevin Dynamics

Abstract

We propose an interacting contour stochastic gradient Langevin dynamics (ICSGLD) sampler, an embarrassingly parallel multiple-chain contour stochastic gradient Langevin dynamics (CSGLD) sampler with efficient interactions. We show that ICSGLD can be theoretically more efficient than a single-chain CSGLD with an equivalent computational budget. We also present a novel random-field function, which facilitates the estimation of self-adapting parameters in big data and obtains free mode explorations. Empirically, we compare the proposed algorithm with popular benchmark methods for posterior sampling. The numerical results show a great potential of ICSGLD for large-scale uncertainty estimation tasks.

Cite

Text

Deng et al. "Interacting Contour Stochastic Gradient Langevin Dynamics." International Conference on Learning Representations, 2022.

Markdown

[Deng et al. "Interacting Contour Stochastic Gradient Langevin Dynamics." International Conference on Learning Representations, 2022.](https://mlanthology.org/iclr/2022/deng2022iclr-interacting/)

BibTeX

@inproceedings{deng2022iclr-interacting,
  title     = {{Interacting Contour Stochastic Gradient Langevin Dynamics}},
  author    = {Deng, Wei and Liang, Siqi and Hao, Botao and Lin, Guang and Liang, Faming},
  booktitle = {International Conference on Learning Representations},
  year      = {2022},
  url       = {https://mlanthology.org/iclr/2022/deng2022iclr-interacting/}
}