A Simple Unified Framework for High Dimensional Bandit Problems

Authors: Wenjie Li, Adarsh Barik, Jean Honorio

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we provide some experiments in order to validate our claims in the theoretical results. We choose to run our algorithms in the different high dimensional bandit problems and validate the corresponding regret upper bounds in Corollary 4.2, Corollary 4.6, Corollary 4.9 and Theorem 5.1. For each problem, we plot the cumulative regret R(T) as well as R(T)/B(T), where B(T) is the upper bound derived in the specific application. For each setting in each problem, we run the algorithm for 10 independent runs and plot the mean results with one standard deviation error bars.
Researcher Affiliation Academia 1Department of Statistics, Purdue University, West Lafayette, IN 2Department of Computer Science, Purdue University, West Lafayette, IN.
Pseudocode Yes Algorithm 1 Explore-the-Structure-Then-Commit
Open Source Code No The paper does not include an unambiguous statement that the authors are releasing their code for the work described in this paper, nor does it provide a direct link to a source-code repository.
Open Datasets No We generate our true θ by randomly choosing its non-zero indices, and then generate each of its non-zero values uniformly randomly from [0,1] and then perform the normalization. We set K = 10 so that there are ten different contexts available at each round. The contexts {xt,ai}K i=1 are generated from the zero-mean and identity-covariance normal distribution.
Dataset Splits No The paper mentions running algorithms for 10 independent runs and plotting mean results, but does not specify any training, validation, or test dataset splits, nor does it mention cross-validation or specific split percentages/counts for the data used in experiments.
Hardware Specification No The paper does not explicitly describe the specific hardware used to run its experiments, such as GPU models, CPU models, or cloud computing instances.
Software Dependencies No The paper does not provide specific software dependencies, such as library names with version numbers, needed to replicate the experiment.
Experiment Setup No The paper provides some setup details like the number of arms (K=10) and data generation methods, but it does not specify concrete hyperparameter values (e.g., learning rate, batch size, number of epochs, optimizer settings) or system-level training configurations.