This code is for the project of the course "Fondements théoriques du deep learning" of master MVA 2020. We propose several numerical experiments to illustrate the methods developped in [1].
We provide notebooks to run experiments:
- Non-linear regression with "Bayesian by backprop"
- Contextual bandit with "Bayesian by backprop" on mushrooms data set
In models you will find all our algorithms that we use for the experiments.
In Pyro you will find a work in progress on an implementation of the "Bayes by backprop" method in Pyro for non-linear regression.
[1] Charles Blundell et al. “Weight Uncertainty in Neural Networks”. In:Proceedings of the 32ndInternational Conference on International Conference on Machine Learning - Volume 37.ICML’15. Lille, France: JMLR.org, 2015, pp. 1613–1622.