Deep Unsupervised Learning using Nonequilibrium Thermodynamics
Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli submitted 'Deep Unsupervised Learning using Nonequilibrium Thermodynamics' on March 12, 2015. The paper describes an iterative forward diffusion process that destroys structure in a data distribution and a learned reverse diffusion process that restores structure.
Historical classification
This verified record belongs to LINEAiGE's artificial-intelligence history layer beginning in 1956.
Why it matters
A documented generative-model research milestone that explicitly formulates learned reverse diffusion for data generation and likelihood-based modeling.
People, organizations & technologies
PEOPLE
Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, Surya Ganguli
TECHNOLOGIES
Diffusion probabilistic models, Deep generative models, Unsupervised learning
Evidence & verification
STATE
verified
CONFIDENCE
high
LAST VERIFIED
2026-09-14T21:16:20+08:00
The original arXiv record establishes the March 12 2015 submission date, title, authors, and the forward/reverse diffusion formulation. The ICML 2015 PMLR proceedings independently corroborate the paper and its publication in Proceedings of Machine Learning Research volume 37. No invention date, first-ever claim, development geography, institutional affiliation, ancestry, influence, causality, relationship edge, or claim about later diffusion systems is inferred.
Sources
Deep Unsupervised Learning using Nonequilibrium Thermodynamics
arXiv
PUBLISHED
2015-03-12
Deep Unsupervised Learning using Nonequilibrium Thermodynamics
Proceedings of Machine Learning Research
Trace the chronology
Record identity
LINEAiGE IDdeep-unsupervised-nonequilibrium-thermodynamics-2015