AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Efficient Symbolic Policy Learning with Differentiable Symbolic Expression

Jiaming Guo, Rui Zhang, Shaohui Peng, Qi Yi, Xing Hu, Ruizhi Chen, Zidong Du, Xishan Zhang, Ling Li, Qi Guo, Yunji Chen

NeurIPS 2023 · Reinforcement learning › Visual and representation RL · approach: Task structure and curricula for meta-RL

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
replace path selector with L1 norm minimization (ablation)ESPL · BipedalWalkeraverage episode reward: 309.43 -> 6.81Hurt
replace path selector with L1 norm minimization (ablation)ESPL · Hopperaverage episode reward: 2442.48 -> 1546.35Hurt
replace symbolic network with plain structure (ablation)ESPL · BipedalWalkeraverage episode reward: 309.43 -> -11.63Hurt
replace symbolic network with plain structure (ablation)ESPL · Hopperaverage episode reward: 2442.48 -> 2003.24Hurt
replace symbolic network with densely connected structure (no arranged operators) (ablation)ESPL · Hopperaverage episode reward: 2442.48 -> 2316.54Hurt
replace symbolic network with densely connected structure (no arranged operators) (ablation)ESPL · Pendulumaverage episode reward: -151.72 -> -144.08Improved
use ESPL instead of symbolic regression from pretrained policy (ablation)regression baseline · CartPoleaverage episode reward: 211.82 -> 1000Improved
use ESPL instead of DSP (ablation)DSP · single-task RLaverage rank (lower is better): 4.625 -> 3.5Improved
use ESPL instead of DSP (training episodes) (ablation)DSP · CartPoleenvironment episodes: 2M -> 500Improved
use ESPL instead of DSP (policy length)DSP · single-task RLaverage symbolic length: 8.25 -> 12.91Hurt
use ESPL instead of TD3TD3 · Hopperaverage episode reward: 2743.9 -> 2442.48Hurt
use CSP symbolic policy instead of PEARL neural policyPEARL · meta-RL deploymentFLOPs: 60-45000xImproved

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