Topology-Aware Revival for Efficient Sparse Training
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866908811659313152 |
|---|---|
| author | Jin, Meiling Wang, Fei Yuan, Xiaoyun Qian, Chen Cheng, Yuan |
| author_facet | Jin, Meiling Wang, Fei Yuan, Xiaoyun Qian, Chen Cheng, Yuan |
| contents | Static sparse training is a promising route to efficient learning by committing to a fixed mask pattern, yet the constrained structure reduces robustness. Early pruning decisions can lock the network into a brittle structure that is difficult to escape, especially in deep reinforcement learning (RL) where the evolving policy continually shifts the training distribution. We propose Topology-Aware Revival (TAR), a lightweight one-shot post-pruning procedure that improves static sparsity without dynamic rewiring. After static pruning, TAR performs a single revival step by allocating a small reserve budget across layers according to topology needs, randomly uniformly reactivating a few previously pruned connections within each layer, and then keeping the resulting connectivity fixed for the remainder of training. Across multiple continuous-control tasks with SAC and TD3, TAR improves final return over static sparse baselines by up to +37.9% and also outperforms dynamic sparse training baselines with a median gain of +13.5%. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_04166 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Topology-Aware Revival for Efficient Sparse Training Jin, Meiling Wang, Fei Yuan, Xiaoyun Qian, Chen Cheng, Yuan Machine Learning Artificial Intelligence Static sparse training is a promising route to efficient learning by committing to a fixed mask pattern, yet the constrained structure reduces robustness. Early pruning decisions can lock the network into a brittle structure that is difficult to escape, especially in deep reinforcement learning (RL) where the evolving policy continually shifts the training distribution. We propose Topology-Aware Revival (TAR), a lightweight one-shot post-pruning procedure that improves static sparsity without dynamic rewiring. After static pruning, TAR performs a single revival step by allocating a small reserve budget across layers according to topology needs, randomly uniformly reactivating a few previously pruned connections within each layer, and then keeping the resulting connectivity fixed for the remainder of training. Across multiple continuous-control tasks with SAC and TD3, TAR improves final return over static sparse baselines by up to +37.9% and also outperforms dynamic sparse training baselines with a median gain of +13.5%. |
| title | Topology-Aware Revival for Efficient Sparse Training |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2602.04166 |