A Distribution Optimization Framework for Confidence Bounds of Risk Measures
Authors: Hao Liang, Zhi-Quan Luo
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We further verify the efficacy of the proposed framework by providing tighter problem-dependent regret bound for the CVa R bandit. ... We validate the proposed framework by applying it to the risk-sensitive bandit problems in Section 6, and provide numerical experiments in Section 7. ... To better visualize the benefits of our framework relative to those of LLC and GLC, we conducted a series of empirical comparisons. Details and complete figures are deferred to Appendix G. ... Figure 4: Comparisons of CIs for CVa R and ERM with varying sample sizes. Figure 5: Cumulative CVa R-regret of CVa R-UCB (red), LLC-UCB (blue), and GLC-UCB (green). |
| Researcher Affiliation | Academia | 1School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen 2Shenzhen Research Institute of Big Data. Correspondence to: Hao Liang <haoliang1@link.cuhk.edu.cn>. |
| Pseudocode | Yes | Algorithm 1 Lower Confidence Bound ... Algorithm 2 Wasserstein upper confidence bound ... Algorithm 3 Wasserstein lower confidence bound ... Algorithm 4 Supremum upper confidence bound ... Algorithm 5 Supremum lower confidence bound |
| Open Source Code | No | The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | We adopt the same bandit instances as Tamkin et al. (2019). The parameters of these distributions are given in Table 1 in Tamkin et al. (2019). |
| Dataset Splits | No | The paper describes the use of 'n i.i.d. samples' and discusses 'sample size' in its experiments, but it does not specify explicit training, validation, or test dataset splits (e.g., percentages or counts) for reproducibility. |
| Hardware Specification | No | The paper does not provide any specific details regarding the hardware used for running the experiments (e.g., GPU/CPU models, memory, cloud resources). |
| Software Dependencies | No | The paper does not list specific software components with version numbers (e.g., programming languages, libraries, frameworks, or solvers) that would be needed to reproduce the experiments. |
| Experiment Setup | Yes | Unless otherwise specified, we always use N = 10^5 samples, α = 0.05, β = 1, and δ = 0.05. For convenience, we use c1 n = (b a)c n . ... We adopt the same bandit instances as Tamkin et al. (2019). |