What Makes a Good Prune?

A simple method to determine the optimal amount by which a network can be L1-pruned based on its parameter distribution

Notebook provided demonstrates the method presented in "What Makes a Good Prune? Maximal Unstructured Pruning for Maximal Cosine Similarity" by G. Mason-Williams and F. Dahlqvist 2024. on the LeNet5 Architecture with the MNIST dataset.

If you use this please cite using:

@inproceedings{
  mason-williams2024what,
  title={What Makes a Good Prune? Maximal Unstructured Pruning for Maximal Cosine Similarity},
  author={Gabryel Mason-Williams and Fredrik Dahlqvist},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2024},
  url={https://openreview.net/forum?id=jsvvPVVzwf }
}