AI Paper Atlas

Research map / Reinforcement learning / MDPs and planning

Identifiability and Generalizability in Constrained Inverse Reinforcement Learning

Andreas Schlaginhaufen, Maryam Kamgarpour

ICML 2023 · Reinforcement learning › MDPs and planning · approach: POMDPs and confounded RL · 2 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 constrained IRL with boundary-only reward features R1 instead of unconstrained IRL with R1Gridworld IRL · Gridworld test (b1 >> b0)occupancy measure error (Δμ): 2.4e-1 -> 1.3e-7Improved
Use constrained IRL with boundary-only reward features R1 instead of unconstrained IRL with R1Gridworld IRL · Gridworld test (b1 >> b0)suboptimality (ΔJ): 2.9e-1 -> 9.2e-14Improved
Use constrained IRL with full state features R2 instead of unconstrained IRL with R2Gridworld IRL · Gridworld test (b1 >> b0)occupancy measure error (Δμ): 2.8e-1 -> 2.1e-2Improved
Use constrained IRL with full state features R2 instead of unconstrained IRL with R2Gridworld IRL · Gridworld test (b1 >> b0)suboptimality (ΔJ): 5.9e-1 -> 1.7e-3Improved
Restrict the reward class to boundary-only features R1 rather than full state features R2 (ablation)Constrained IRL · Gridworld test (b1 >> b0)occupancy measure error (Δμ): 2.1e-2 -> 1.3e-7Improved
Restrict the reward class to boundary-only features R1 rather than full state features R2 (ablation)Constrained IRL · Gridworld test (b1 >> b0)suboptimality (ΔJ): 1.7e-3 -> 9.2e-14Improved
Use boundary-only features R1 rather than full state features R2 (ablation)Unconstrained IRL · Gridworld test (b1 >> b0)occupancy measure error (Δμ): 2.8e-1 -> 2.4e-1Improved
Use boundary-only features R1 rather than full state features R2 (ablation)Unconstrained IRL · Gridworld test (b1 >> b0)suboptimality (ΔJ): 5.9e-1 -> 2.9e-1Improved

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