Tensorflow implementation of Bayes by Backprop from the following paper "Weight Uncertainty in Neural Networks" available at https://arxiv.org/pdf/1505.05424.pdf.
The implementation is inspired from:-
- https://gluon.mxnet.io/chapter18_variational-methods-and-uncertainty/bayes-by-backprop.html, and
- https://github.com/christegho/bnn-mnist.git
Implementation does not use OOP principles as the objective here is to have a basic understandable and functional code for the algorithm.
The MNIST dataset is used, as in the paper. Although implementation 1. (above) suggests that n_samples parameter=1 works, it didn't work for me. I had to use n_samples=5 for decent results. Please let me know if you have any feedback for the implementation via. email: [email protected]. Thanks!