We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: MCMC Should Mix: Learning Energy-Based Model with Neural Transport Latent Space MCMC

Abstract: Learning energy-based model (EBM) requires MCMC sampling of the learned model as an inner loop of the learning algorithm. However, MCMC sampling of EBMs in high-dimensional data space is generally not mixing, because the energy function, which is usually parametrized by a deep network, is highly multi-modal in the data space. This is a serious handicap for both theory and practice of EBMs. In this paper, we propose to learn an EBM with a flow-based model (or in general a latent variable model) serving as a backbone, so that the EBM is a correction or an exponential tilting of the flow-based model. We show that the model has a particularly simple form in the space of the latent variables of the backbone model, and MCMC sampling of the EBM in the latent space mixes well and traverses modes in the data space. This enables proper sampling and learning of EBMs.
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:2006.06897 [stat.ML]
  (or arXiv:2006.06897v2 [stat.ML] for this version)

Submission history

From: Erik Nijkamp [view email]
[v1] Fri, 12 Jun 2020 01:25:51 GMT (9068kb,D)
[v2] Wed, 16 Mar 2022 07:53:07 GMT (9340kb,D)

Link back to: arXiv, form interface, contact.