 NTK-SAP: Improving neural network pruning by aligning training dynamicsInternational Conference on Learning Representations (ICLR), 2023 |
 When Expressivity Meets Trainability: Fewer than Neurons Can WorkNeural Information Processing Systems (NeurIPS), 2022 |
 Early-stopped neural networks are consistentNeural Information Processing Systems (NeurIPS), 2021 |