AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

An Off-policy Policy Gradient Theorem Using Emphatic Weightings

Ehsan Imani, Eric Graves, Martha White

NeurIPS 2018 · Reinforcement learning › Policy gradient and actor-critic · approach: Trajectory reuse in policy gradients · 4 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
Add emphatic state weighting to the policy gradientACE versus OffPAC · Three-state aliased MDPobjectiveImproved
Use the ordinary behavior-state weighting instead of emphatic weighting (λa = 0) (ablation)ACE · Three-state aliased MDPobjectiveHurt
Set λa to 0.25 rather than using full emphatic weighting (ablation)ACE · Three-state aliased MDPobjectiveMixed
Use λa = 0.5 rather than λa = 0 (ablation)ACE · Three-state aliased MDPobjectiveImproved
Use lower λa values (ablation)ACE · Three-state aliased MDPlearning speedMixed
Use a GTD(λ) critic rather than exact value estimates (ablation)ACE · Three-state aliased MDPobjectiveMixed
Estimate emphatic weightings online rather than using exact current-policy weightings (ablation)ACE · 11-state aliased MDPobjectiveHurt
Add true emphatic weighting to DPGDPG · Continuous-action MDPperformanceImproved
Use estimated emphatic weightings in ACE rather than the exact weighting (ablation)ACE · Continuous-action MDPperformanceHurt
Use a stochastic policy rather than a deterministic policy (ablation)ACE versus True-DPGE · Continuous-action MDPperformanceHurt

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