Skip to content

alexrakowski/mc-dropout-mnist

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 

Repository files navigation

mc-dropout-mnist

Implementation of (parts of) the experiment on MNIST from Bayesian Convolutional Neural Networks with Bernoulli Approximate Variational Inference

Standard LeNet architecture without Dropout is compared against a LeNet-all architecture, where Dropout is applied after each layer (including convolutions). Dropout is kept at test time, and the prediction of the trained model is averaged over T=50 stochastic passes. The MC-Dropout model achieves an error rate of ~0.6%, compared to ~1% of the non-dropout model.

Required libraries: tqdm, keras

Tested with Tensorflow and Python 3.