Offline Inverse RL: New Solution Concepts and Provably Efficient Algorithms
Authors: Filippo Lazzati, Mirco Mutti, Alberto Maria Metelli
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper, we introduce a novel notion of feasible reward set capturing the opportunities and limitations of the offline setting and we analyze the complexity of its estimation. Then, we propose two computationally and statistically efficient algorithms, IRLO and PIRLO, for addressing the problem. We present two provably efficient algorithms, IRLO and PIRLO. We have applied PIRLO to the highway driving application domain. To this aim, we have used the data gathered by Likmeta et al. (2021). |
| Researcher Affiliation | Academia | 1Politecnico di Milano, Milan, Italy 2Technion, Haifa, Israel. Correspondence to: Filippo Lazzati <filippo.lazzati@polimi.it>. |
| Pseudocode | Yes | Algorithm 1 IRLO and PIRLO. Algorithm 2 Membership checker for IRLO and PIRLO. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | Yes | We have applied PIRLO to the highway driving application domain. To this aim, we have used the data gathered by Likmeta et al. (2021). The data is publicly available at https://github.com/amarildolikmeta/irl_real_life/tree/main/datasets/highway. |
| Dataset Splits | No | The paper describes the collection of two datasets, `Db` and `DE`, but does not specify standard training, validation, and test splits for model evaluation. |
| Hardware Specification | No | The paper does not provide specific hardware details used for running its experiments. |
| Software Dependencies | No | The paper mentions software components but does not provide specific version numbers for reproducibility (e.g., Python, PyTorch, CUDA, etc.). |
| Experiment Setup | No | The paper describes data preprocessing and experiment design (e.g., discrete features, state space size), but lacks specific hyperparameters or system-level training settings for the algorithms themselves. |