/tf-bayes-by-backprop

Tensorflow implementation of the Bayes by Backprop algorithm as proposed by Blundell et al. in "Weight Uncertainty in Neural Networks" (2015) - https://arxiv.org/pdf/1505.05424.pdf

Primary LanguageJupyter Notebook

Tensorflow - Bayes by Backprop

Description

This repository contains a basic implementation of the "Weight uncertainty in neural networks" paper for approximate variational inference by Blundell et al. using TensorFlow. In order to show the possibilities of the network, the code is accompanied by a python notebook with a simple regression problem.

img

Prerequisites

Minimal Requirements:

  • Tensorflow 2.x
  • TensorflowProbability
  • overrides

Additional Requirements:

Note: The following python packages are only required for running the accompanying code example in _example.ipynb

  • numpy
  • matplotlib

References

[1] Charles Blundell, Julien Cornebise, Koray Kavukcuoglu, and Daan Wierstra. 2015. Weight uncertainty in neural networks. In Proceedings of the 32nd International Conference on International Conference on Machine Learning - Volume 37 (ICML'15). JMLR.org, 1613–1622.