AI Paper Atlas

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

Lifelong Policy Gradient Learning of Factored Policies for Faster Training Without Forgetting

Jorge Mendez, Boyu Wang, Eric Eaton

NeurIPS 2020 · Reinforcement learning › Policy gradient and actor-critic · approach: Trajectory reuse in policy gradients · 11 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
learn each new task's coefficients by policy gradient in the span of a shared factor dictionaryLPG-FTW versus STL · Six MuJoCo domainslearning speedImproved
learn each new task's coefficients by policy gradient in the span of a shared factor dictionaryLPG-FTW versus STL · Five of six MuJoCo domainsfinal performanceImproved
update the shared factor dictionary using the current task's informationLPG-FTW · Six MuJoCo domainsaverage return after dictionary updateImproved
retain a shared factor dictionary and task-specific coefficients rather than train isolated policiesLPG-FTW versus PG-ELLA · Six MuJoCo domainscatastrophic forgettingImproved
use a shared single policy across tasksEWC · MuJoCo domainslearning speedMixed
use a shared single policy across tasksEWC · Two MuJoCo domainsperformance relative to STLHurt
learn each new task's coefficients in the shared factor spanLPG-FTW versus baselines · Meta-World MT10learning speedImproved
learn each new task's coefficients in the shared factor spanLPG-FTW versus baselines · Meta-World MT50learning speedImproved
use the shared factor dictionary after sequential trainingLPG-FTW · Meta-World MT10catastrophic forgettingMixed
use the shared factor dictionary after sequential trainingLPG-FTW · Meta-World MT50catastrophic forgettingImproved
increase EWC network capacity (ablation)EWC · Meta-World MT10 and MT50learning and forgetting performanceNo clear effect
initialize dictionary columns with separate single-task solutions instead of incremental error vectors (ablation)LPG-FTW initialization · Early sequential taskslearning speedHurt

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