AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Efficient Probabilistic Performance Bounds for Inverse Reinforcement Learning

Daniel Brown, Scott Niekum

AAAI 2018 · Reinforcement learning › Imitation learning · approach: Reward learning from demonstrations · 36 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 posterior-sampled VaR bounds instead of worst-case feature-count boundsPolicy evaluation from demonstrations · Grid navigationbound error: often four times tighterImproved
Use the 0.95-VaR EVD bound instead of the Hoeffding-style EVD boundProjection algorithm · Grid navigationaverage EVD bound: 0.9372 vs 0.9372; 23,146 demonstrations vs 1 demonstrationImproved
Increase BIRL confidence parameter c on noisy demonstrations (ablation)0.95-VaR bound · Noisy grid navigationbound errorMixed
Set BIRL confidence parameter c to 50 for noisy demonstrations (ablation)0.95-VaR bound · Noisy grid navigationaccuracyHurt
Use BIRL confidence parameter c in the range (1,10] for noisy demonstrations (ablation)0.95-VaR bound · Noisy grid navigationaccuracyImproved
Increase VaR quantile from 0.95 to 0.99 (ablation)EVD confidence bound · Grid navigationaccuracy: 0.98 -> 1.0Improved
Increase VaR quantile from 0.95 to 0.99 (ablation)EVD confidence bound · Grid navigation, 1 demonstrationaverage EVD bound: 0.9372 -> 1.1428Hurt
Use the VaR bound rather than WFCB to rank driving policiesPolicy selection · Drivingpolicy ranking: right-safe: 3 -> 1; on-road: 1 -> 2; nasty: 2 -> 3Improved
Optimize the 0.99-VaR bound by changing policy actions one state at a timeBIRL maximum-likelihood policy · Two-feature navigation0.99-VaRImproved
Use EVD instead of feature-count discrepancy to evaluate policiesRisk-aware policy bound · Partial or noisy demonstrationsbound applicabilityImproved

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