A Tensorflow implementation of bayes-by-backprop from Weight uncertainty in neural networks paper (https://arxiv.org/abs/1505.05424).
It was a homework assignment again, so do not expect a nice code. Note that this code may have differences (in terms of parameters, constants and use of variables) from other implementations of the same code, since I followed the given assignment pdf.
As a final note, you may need to download and/or change the input path in "cifar_input.py", if you want to run this code directly.