AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

The State of Sparse Training in Deep Reinforcement Learning

Laura Graesser, Utku Evci, Erich Elsen, Pablo Samuel Castro

ICML 2022 · Architectures and efficiency › Pruning and efficient tuning · approach: Sparse training from scratch · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use sparse training rather than a parameter-equivalent dense networkDQN, SAC, and PPO agents · Atari and MuJoCorewardImproved
Train sparse networks with pruning instead of dense trainingDQN, SAC, and PPO agents · Atari and MuJoCorewardImproved
Use dynamic sparse training instead of static sparse trainingDQN, SAC, and PPO agents · Atari and MuJoCorewardImproved
Use RigL instead of SETDQN, SAC, and PPO agents · Atari and MuJoCorewardNo clear effect
Use sparse networks at 80-90% sparsityDQN and SAC agents · Atari and MuJoCorewardNo clear effect
Use a ResNet instead of the standard CNNDQN · AtariIQM score: about 3x improvement in IQM scoresImproved
Allocate more parameters to the critic and fewer to the actor (ablation)SAC · MuJoCoreturnImproved
Use ERK instead of uniform sparsity distribution (ablation)Sparse DQN, SAC, and PPO networks · Atari and MuJoCorewardImproved
Add a small amount of weight decay (ablation)Pruning, RigL, and SET · MuJoCorewardImproved
Use sparsity-aware initialization with uniform layer sparsity (ablation)Static, RigL, and SET · MuJoCorewardImproved
Use sparsity-aware initialization with ERK sparsity (ablation)RigL and SET · MuJoCorewardNo clear effect
Use sparsity-aware initialization with ERK sparsity (ablation)Static · MuJoCorewardHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy