[KongchangAI]
ConceptEBMs / Energy-Based Models

Energy-Based Models

A class of probabilistic models that assign a scalar energy value to each state, with low energy indicating high probability or high fitness. They do not require the normalization constant to be computable and can flexibly represent complex distributions.

Timeline (last 90 days)

Nov 18

EBMs usually only require defining an energy function, and the partition function is difficult to compute, so training and sampling often rely on approximations such as contrastive divergence, MCMC, or Langevin dynamics.

Unverified50%
Nov 18

EBMs face problems such as training instability, mode collapse, and high sampling costs.

Unverified50%
Nov 9

Energy-based models (EBMs) are a class of probabilistic models that assign a value to each state via a scalar energy function, with low energy indicating high probability or high fitness, and that do not require the normalization constant to be computable.

Unverified50%
Sep 12

Energy-based models are a class of framework that Yann LeCun is optimistic about.

Unverified50%

All Facts (4)

Source Articles