SparseProp: Efficient Sparse Backpropagation for Faster Training of Neural Networks

Published in ICML 2023, 2023

My groupmates found a way to do faster backpropagation through unstructured-sparsity weights. I helped with the experiments.

arXiv