Finite Time Logarithmic Regret Bounds for Self-Tuning Regulation
Authors: Rahul Singh, Akshay Mete, Avik Kar, Panganamala Kumar
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comparative simulation results illustrate the improved performance of PIECE. In Section 5 we present the comparative results obtained in simulation experiments. |
| Researcher Affiliation | Academia | 1Indian Institute of Science, Bengaluru, Karnataka, India. 2Texas A&M University, College Station, TX, USA. |
| Pseudocode | Yes | Algorithm 1 PIECE: Probing Inputs for Exploration in Certainty Equivalence. Algorithm 2 Certainty Equivalence (CE). Algorithm 3 Lai and Wei (LW). |
| Open Source Code | No | The paper does not provide any concrete access information, such as a repository link or an explicit statement of code release, for the methodology described. |
| Open Datasets | Yes | EXAMPLE I (PAPER MACHINE (ÅSTRÖM & WITTENMARK, 1973)): Linear system with p = 2 and q = 2: yt = 1.283yt 1 0.495yt 2+2.307ut 1 2.025ut 2+wt. EXAMPLE II: An linear system with p = 4 and q = 4: yt = 1.18yt 1 0.48yt 2 +0.45yt 3 0.41yt 4 + 0.28ut 1 + 0.14ut 2 + 0.16ut 3 + 0.03ut 4 + wt. |
| Dataset Splits | No | The paper describes simulation experiments on defined ARX systems rather than using pre-split datasets. There is no mention of specific training, validation, or test dataset splits, percentages, or sample counts for reproducibility. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as exact GPU/CPU models, processor types, or memory specifications used for running the simulations. |
| Software Dependencies | No | The paper does not list specific software components with their version numbers that would be necessary to reproduce the experiments. While simulations are mentioned, no particular software dependencies are explicitly detailed. |
| Experiment Setup | Yes | Each simulation experiment is performed for 1000 steps. The reported results are the averaged values over 50 runs. Hyper-Parameters: PIECE needs two system-dependent parameters and a bound on the absolute value of the noise in order to compute the algorithm s hyper-parameters. ρ is an upper bound on the eigenvalues of matrix A (see (2)) and λ 2 is the ℓ2-norm of vector λ. The duration of the first exploration episode is λ 3 2... Bw is the upper bound for the absolute value of the noise sequence. Bu, the threshold for the clipping input... H, defined as in (12), is the exploration episode duration after the first episode. The paper also includes tables like "Table 4. System parameters" and "Table 6. PIECE hyper-parameters for Gaussian noise with mean 0 and standard deviation 0.2." with specific parameter values. |