INNF+ 2020

This workshop focuses on the design of probabilistic models that are flexible yet remain tractable for general usage. While research started with heavily constrained models, such as the use of Gaussian assumptions or structured dependency assumptions, recent progress is moving towards universal high-dimensional density models. Remaining tractable (in terms of efficient likelihood evaluation, sampling, and/or marginalization) allows simple plug-and-play applications of such probabilistic models in a broad variety of domains and scientific fields. Examples include: hierarchical reinforcement learning [1], Bayesian deep learning [2], PAC-Bayes generalization bound estimation [3], likelihood-free inference [4], lossless compression [5], probabilistic programming [6, 7], biomedics [8] and statistical physics [9].

Submission Details

We invite researchers to submit their recent progress on the development, analysis, or application of work related to likelihood-based generative models and invertible neural networks. Submissions should take the form of an extended abstract of 4 pages in PDF format using the ICML style. Submissions longer than 4 pages are allowed but reviewers are not expected to read more than 4 pages. Author names do not need to be anonymized. Submissions may include a supplementary/appendix, but reviewers are not responsible for reading any supplementary material. Submissions that are currently under review or that have been recently accepted for publication to another conference are permitted.

Potential topics include but are not limited to:

  • Normalizing flows and other likelihood-based generative modeling
  • Proposing new structural constraints to construct likelihood-based models
  • Designing invertible transformations parameterized by neural networks
  • Improving approximate inference
  • Theoretical work in terms of optimization and/or expressivity of invertible networks
  • Continuous-time normalizing flows and ODEs
  • Reinforcement learning with density models
  • Normalizing flows with discrete distributions
  • Continuous relaxation to discrete variables
  • Probabilistic programming

Please submit here. The submission deadline will be May 15th June 5th June 10th, at 23h59 anywhere on earth.

Questions? Contact us at