By Kevin Ammouri and Youssef Taoudi
Project aiming to replicate The Lottery Ticket Hypothesis by J. Frankle and M. Carbin (https://arxiv.org/abs/1803.03635)
Winning tickets were examined using the LeNet-300-100(LeCun) architecture on the MNIST dataset and miniature versions of the VGG model(Simonyan,http://arxiv.org/abs/1409.1556) for CIFAR-10.
conv_cifar consists all code, data and plots for the convolutional network pruning while fc_mnist contain all code, data and graphs for the fully connected network pruning.
constants.py - architecture and hyperparameter setup
experiment.py code for running the LeNet experiments
conv_models.py - code for creating and training the models as well as applying the mask for pruning
plots.py - code for plotting
pruning.py - code for pruning
tools.py - helper functions
constants.py - architecture and hyperparameter setup
conv_experiment.py - code for running the ConvNet experiments
conv_models.py - code for creating and training the models as well as applying the mask for pruning
plots.py - code for plotting
pruning.py - code for pruning
tools.py - helper functions