durga

HRel: Filter Pruning based on High Relevance between Activation Maps and Class Labels

.yaml file of the conda environment is also included

Acknowledgements

Code for Mutual information estimation and Information plane analysis is taken from Wickstrom

Few blocks of HRank code are used

Running the code

Run the file with the suffix 'ex.py' in the respective architectures.

At the end '.npz ' file is created with the Mutual information and sigma values for plotting.

Citation

@article{sarvani2022hrel, title={HRel: Filter pruning based on High Relevance between activation maps and class labels}, author={Sarvani, CH and Ghorai, Mrinmoy and Dubey, Shiv Ram and Basha, SH Shabbeer}, journal={Neural Networks}, volume={147}, pages={186--197}, year={2022}, publisher={Elsevier} }