AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Hyperparameters in Reinforcement Learning and How To Tune Them

Theresa Eimer, Marius Lindauer, Roberta Raileanu

ICML 2023 · Reinforcement learning › Visual and representation RL · approach: Probabilistic and autonomous RL · 9 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 HPO rather than manual hyperparameter sweepsRL agents · Across tested RL environmentscompute budget for comparable or better configurations: less than 1/12 of the budgetImproved
Tune the full search space rather than only learning rate (ablation)PPO on Acrobot · Acrobotnegative evaluation reward: 71 ± 1 -> 71 ± 3No clear effect
Expand the search space from learning rate only to full (ablation)SAC on Pendulum · Pendulumnegative evaluation reward: 71 ± 12 -> 112 ± 24Hurt
Expand the search space from learning rate only to full (ablation)PPO on Acrobot · Acrobotnegative evaluation reward: 94 ± 22 -> 305 ± 186Hurt
Expand the search space from learning rate only to full (ablation)SAC on Pendulum · Pendulumnegative evaluation reward: 207 ± 126 -> 78 ± 19Improved
Increase tuning seeds from 1 to 5 (ablation)DEHB · Acrobottest reward: 341.3±183.1 -> 465.1±24.6Improved
Increase tuning seeds from 1 to 5 (ablation)DEHB · Pendulumtest reward: 150.5±13.4 -> 350.2 ± 418.2Mixed
Use separate test seeds rather than judging only by tuning-seed incumbentsHPO-selected RL configurations · Acrobot and Pendulumperformance generalization across seedsHurt
Use DEHB instead of hand-tuned baseline configurationsPPO on Brax · Braxmean rank: 1.7 -> 1.3 (16 runs); 1.3 -> 1.0 (64 runs)Improved
Use DEHB instead of hand-tuned baseline configurationsIDAAC on Procgen · Procgenmean rank: 2 -> 1.7 (16 runs); 1.3 -> 1.0 (64 runs)Improved
Increase RS tuning budgetRL agents on Brax and Procgen · Brax and Procgenmean rank: 2.3 -> 3.3 (Brax); 2.7 -> 3 (Procgen)Hurt
Tune across more seeds, beyond 3 (ablation)HPO methods · Acrobot and Pendulumtest performanceMixed

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