Stage-wise Conservative Linear Bandits
Authors: Ahmadreza Moradipari, Christos Thrampoulidis, Mahnoosh Alizadeh
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we investigate the numerical performance of SCLTS and SCLUCB on synthetic data, and compare it with SEGE algorithm introduced by Khezeli and Bitar (2019). In all the implementations, we used the following parameters: R = 0.1, S = 1, a = 1, d = 2. We consider the action set X to be a unit ball centered on the origin. The reward parameter T is drawn from N(0, I4). We generate the sequence { bt}T t=1 to be IID random vectors that are uniformly distributed on the unit circle. The results are averaged over 100 realizations. In Figure 1(left), we plot the cumulative regret of the SCLTS algorithm and SCLUCB and SEGE algorithm from Khezeli and Bitar (2019) for = 0.2 over 100 realizations. |
| Researcher Affiliation | Academia | Ahmadreza Moradipari, Christos Thrampoulidis, Mahnoosh Alizadeh Department of Electrical and Computer Enginnering University of California, Santa Barbara ahmadreza_moradipari@ucsb.edu |
| Pseudocode | Yes | Algorithm 1: Stage-wise Conservative Linear Thompson Sampling (SCLTS) |
| Open Source Code | No | The paper does not provide any explicit statement about making the source code for its methodology publicly available, nor does it provide a link to a code repository or mention code in supplementary materials. |
| Open Datasets | No | In this section, we investigate the numerical performance of SCLTS and SCLUCB on synthetic data, and compare it with SEGE algorithm introduced by Khezeli and Bitar (2019). ... The reward parameter T is drawn from N(0, I4). We generate the sequence { bt}T t=1 to be IID random vectors that are uniformly distributed on the unit circle. The data is generated internally, not from a public source. |
| Dataset Splits | No | The paper mentions generating 'synthetic data' and averaging results 'over 100 realizations', but does not specify any training, validation, or test dataset splits, cross-validation methods, or other data partitioning details. |
| Hardware Specification | No | The paper discusses numerical performance on synthetic data and mentions parameters used for implementation (e.g., R, S, a, d), but it does not provide any specific hardware details such as GPU/CPU models, memory, or cloud computing instances used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies or version numbers for any libraries, frameworks, or solvers used in the implementation of the algorithms. |
| Experiment Setup | Yes | In all the implementations, we used the following parameters: R = 0.1, S = 1, a = 1, d = 2. We consider the action set X to be a unit ball centered on the origin. The reward parameter T is drawn from N(0, I4). We generate the sequence { bt}T t=1 to be IID random vectors that are uniformly distributed on the unit circle. The results are averaged over 100 realizations. ... for different values of over a horizon T = 10000. |