This repository contains the implementation to replicate the experiments "Lottery Ticket Hypothesis". The detail explanation of the concepts are described in this site

THE LOTTERY TICKET HYPOTHESIS: FINDING SPARSE, TRAINABLE NEURAL NETWORK

pip install -r requirements.txt

Cifar-10

VGG-based 6 Convolutional Netowrok and 2 Full Connected Network.The trained model reached 84.82% accuracy on testing set. Detail parameter configs and performance described in [to be linked!].

11.60% accuracy with original data, on the other hand, a pruned accuracy reached 19.10%. However it is found that even not so much rate of pruning can decrease accuracy. Detail result described in [to be linked!].

T.B.C

In root directory of this repository,

python experiments\original_perfoemance\original_performance.py

In root directory of this repository,

python experiments\original_perfoemance\original_performance.py

In root directory of this repository,

python experiments\original_perfoemance\retrain.py