# andrewgordonwilson/bayesgan

Tensorflow code for the Bayesian GAN (https://arxiv.org/abs/1705.09558) (NIPS 2017)

Repository: https://github.com/andrewgordonwilson/bayesgan
Canonical: https://ross.abutalabs.com/products/bayesgan
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Last push: 2018-07-30T20:50:23+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3225, "days_push": 2956, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1013, forks 172 (observed 2026-08-28T04:03:13.778399+00:00)

## What it is
TensorFlow implementation of the Bayesian GAN from the NIPS 2017 paper by Saatchi and Wilson, which places conditional posteriors over generator and discriminator weights and marginalizes them via stochastic gradient Hamiltonian Monte Carlo. It includes scripts for synthetic data and standard image datasets like MNIST, CIFAR10, CelebA, and SVHN.

## Use cases
- reproduce the Bayesian GAN paper results
- train a GAN that avoids mode collapse
- semi-supervised learning with generative adversarial networks
- sample from a probabilistic ensemble of generative models
- run GAN experiments on MNIST or CelebA
- research Bayesian approaches to deep generative models

## When to choose
- you want the reference implementation of the Bayesian GAN for research or paper reproduction
- you need a GAN formulation with uncertainty over weights and resistance to mode collapse
- you are studying Bayesian deep learning methods

## When to avoid
- you need modern, maintained code - it requires Python 2.7 and TensorFlow 1.0.0
- you want production-ready GAN training with current frameworks
- you need a general-purpose GAN library rather than a single paper's code

## Facets
- artifact type: library
- maturity: abandoned
- function: machine-learning, deep-learning, data-generation
- domain: machine-learning, deep-learning, artificial-intelligence
- platform: python
- tags: gan, bayesian-deep-learning, tensorflow, research-code, nips-2017, linux

## Member repositories
- andrewgordonwilson/bayesgan (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:13.778399+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:11:38.139943+00:00, confidence not recorded.
  - readme: https://github.com/andrewgordonwilson/bayesgan (fetched 2026-08-28T04:03:13.778399+00:00, sha b9981f8f6e20)
- Data as of 2026-08-30T08:39:29.467469+00:00.