https://github.com/christopher-beckham/coms-are-energy-models
Official code for paper: Conservative objective models are a special kind of contrastive divergence-based energy model
https://github.com/christopher-beckham/coms-are-energy-models
coms conservative-objective-models contrastive-divergence energy-models generative-design generative-models model-based-optimisation model-based-optimization offline-rl pytorch
Last synced: 3 months ago
JSON representation
Official code for paper: Conservative objective models are a special kind of contrastive divergence-based energy model
- Host: GitHub
- URL: https://github.com/christopher-beckham/coms-are-energy-models
- Owner: christopher-beckham
- Created: 2023-03-11T20:35:46.000Z (over 3 years ago)
- Default Branch: master
- Last Pushed: 2023-08-15T15:49:33.000Z (almost 3 years ago)
- Last Synced: 2025-04-06T07:06:58.719Z (about 1 year ago)
- Topics: coms, conservative-objective-models, contrastive-divergence, energy-models, generative-design, generative-models, model-based-optimisation, model-based-optimization, offline-rl, pytorch
- Language: Jupyter Notebook
- Homepage: https://arxiv.org/abs/2304.03866
- Size: 5.95 MB
- Stars: 14
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.org
Awesome Lists containing this project
README
* coms-are-energy-models
Official code for the paper: *Conservative objective models are a special kind of contrastive divergence-based energy model*
#+begin_quote
In this work we theoretically show that conservative objective models (COMs) for offline model-based optimisation (MBO) are a special kind of contrastive divergence-based energy model, one where the energy function represents both the unconditional probability of the input and the conditional probability of the reward variable. While the initial formulation only samples modes from its learned distribution, we propose a simple fix that replaces its gradient ascent sampler with a Langevin MCMC sampler. This gives rise to a special probabilistic model where the probability of sampling an input is proportional to its predicted reward. Lastly, we show that better samples can be obtained if the model is decoupled so that the unconditional and conditional probabilities are modelled separately.
#+end_quote
[[./assets/coms-animation-compressed.gif]]
* Reproduction
All experiments are self-contained in the IPython notebook file!