https://github.com/gd-zhang/differentiable-ais

Implementation for DAIS
https://github.com/gd-zhang/differentiable-ais

Last synced: about 1 month ago
JSON representation

Implementation for DAIS

• Host: GitHub
• URL: https://github.com/gd-zhang/differentiable-ais
• Owner: gd-zhang
• Created: 2021-12-21T20:05:35.000Z (almost 3 years ago)
• Default Branch: main
• Last Pushed: 2021-12-21T20:12:42.000Z (almost 3 years ago)
• Last Synced: 2023-08-01T22:58:07.876Z (over 1 year ago)
• Language: Python
• Size: 17.6 KB
• Stars: 6
• Watchers: 2
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Differentiable Annealed Importance Sampling (DAIS)

This repository contains the code to reproduce the DAIS results from the paper [Differentiable Annealed Importance Sampling and the Perils of Gradient Noise](https://openreview.net/forum?id=6rqjgrL7Lq).

## Bayesian Linear Regression

Please first install JAX (https://github.com/google/jax), then reproduce our results by running the following:

```

python blr/dais.py --gamma 0.0

python blr/dais.py --gamma 0.9

python blr.dais.py --bsize 100

```

## Variational Autoencoder

Please first install PyTorch (https://pytorch.org/).

For training, one example command for DAIS with $K = 10$ and $S = 5$ (to adapt annealing scheme, add `--adapt_beta`):

```

python vae/mnist_train.py --lf_step 10 --n_particles 5 --lf_lrate 0.08

```

After training, you can find the saved checkpoint and evaluate it (e.g., using AIS):

```

python vae/mnist_eval.py --ais --n_particles 10 --lf_step 10000 --lf_lrate 0.05 --resume xxx

```

Also, you can reproduce the evaulation results of AIS/HAIS/DAIS by runing `vae/mnist_eval_scaling.py`:

```

python vae/mnist_eval_scaling.py --linear_beta --lf_lrate 0.08 --n_particles 10 --resume xxx

```

## Citation

To cite this work, please use

```

@inproceedings{

	zhang2021differentiable,

	title={Differentiable Annealed Importance Sampling and the Perils of Gradient Noise},

	author={Guodong Zhang and Kyle Hsu and Jianing Li and Chelsea Finn and Roger Baker Grosse},

	booktitle={Thirty-Fifth Conference on Neural Information Processing Systems},

	year={2021},

	url={https://openreview.net/forum?id=6rqjgrL7Lq}

}

```