23
https://pubmed.ncbi.nlm.nih.gov/38113719
This study presents a new neural network compression scheme using ℓ(0)-norm-based regularization and weight pruning to reduce deployment costs while maintaining inference accuracy.