Pruning Deep Neural Networks via the Marchenko--Pastur Distribution

In three linesNeural network pruning method using Marchenko–Pastur random-matrix theory with minimal post-pruning fine-tuning. On ImageNet-1k, ViT-B/16 achieves 83.41% top-1 with 59.81% MAC reduction after 3 distillation epochs; ResNet50 8:16 reaches 75.87% with 1.62× A40 speedup.

Summary generated by Claude — human-verified