Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Regret Bounds for Episodic Risk-Sensitive Linear Quadratic Regulator
Authors: Wenhao Xu, Xuefeng Gao, Xuedong He
ICLR 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper, we propose a simple least-squares greedy algorithm and show that it achieves e O(log N) regret under a specific identifiability assumption... Our proof relies on perturbation analysis... 5 SIMULATION STUDIES We perform simulation studies to illustrate the regret performances of Algorithms 1 and 2. |
| Researcher Affiliation | Academia | Department of Systems Engineering and Engineering Management The Chinese University of Hong Kong, Hong Kong, China EMAIL |
| Pseudocode | Yes | Algorithm 1 The Least-Squares Greedy Algorithm Algorithm 2 The Least-Squares-Based Algorithm with Exploration Noise |
| Open Source Code | No | The paper does not explicitly state that source code for the methodology described is publicly available, nor does it provide a link to a code repository. |
| Open Datasets | No | The paper describes generating system matrices and parameters for 'System 1', 'System 2', and 'System 3' for simulation studies. It does not use or provide access to any external publicly available datasets. For System 1, it mentions 'We use the system matrices and cost matrices in Section 6.1 of Dean et al. (2020)', which refers to a source for generating part of the system, not a public dataset used for experiments. |
| Dataset Splits | No | The paper conducts simulation studies based on generated system parameters, not on external datasets that would typically require training/test/validation splits. Therefore, no dataset split information is provided. |
| Hardware Specification | Yes | Our experiments are conducted on a PC with 2.10 GHz Intel Processor and 16 GB of RAM. |
| Software Dependencies | No | The paper does not provide specific software names with version numbers (e.g., Python, PyTorch, MATLAB, or specific libraries/solvers with versions) used for implementation or experimentation. |
| Experiment Setup | Yes | In both algorithms, we randomly generate the initial guess θ1 = (A1, B1) with all entries of A1 and B1 sampled from the uniform distribution. In Algorithm 1, we set m1 = 500 and L = l log2 N m1 + 1 m . In Algorithm 2, we set λ = 0.8. System 1. We use the system matrices and cost matrices in Section 6.1 of Dean et al. (2020) with the following minor change: we set QT = Q = 1/2I instead of Q = 10^-3 I as in their paper because the effect of risk parameters is difficult to visualize when Q has small eigenvalues. |