ICML 2025 Oral Paper
Congratulations to our co-founder, Guozheng Ma, for having his article accepted as an oral article by 2025 International Conference on Machine Learning (ICML)!
The ICML is a leading international academic conference in machine learning. Along with NeurIPS and ICLR, it is one of the three most respected conferences of high impact in machine learning and artificial intelligence research.
- Title:
Network Sparsity Unlocks the Scaling Potential of Deep Reinforcement Learning
- Abstract:
Effectively scaling up deep reinforcement learning models has proven notoriously difficult due to network pathologies during training, motivating various targeted interventions such as periodic reset and architectural advances such as layer normalization. Instead of pursuing more complex modifications, we show that introducing static network sparsity alone can unlock further scaling potential beyond their dense counterparts with state-of-the-art architectures. This is achieved through simple one-shot random pruning, where a predetermined percentage of network weights are randomly removed once before training. Our analysis reveals that, in contrast to naively scaling up dense DRL networks, such sparse networks achieve both higher parameter efficiency for network expressivity and stronger resistance to optimization challenges like plasticity loss and gradient interference. We further extend our evaluation to visual and streaming RL scenarios, demonstrating the consistent benefits of network sparsity.
- Code: