Bandits for BMO Functions

Authors: Tianyu Wang, Cynthia Rudin

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We deploy Algorithms 1 and 2 on the Himmelblau’s function and the Styblinski-Tang function (arm space normalized to [0, 1)2, function range rescaled to [0, 10]). The results are in Figure 3. We measure performance using traditional regret and δ-regret. Traditional regret can be measured because both functions are continuous, in addition to being BMO.
Researcher Affiliation Academia 1Department of Computer Science, Duke University, Durham, NC, USA. Correspondence to: Tianyu Wang <tianyu@cs.duke.edu>.
Pseudocode Yes Algorithm 1 Bandit-BMO-Partition (Bandit-BMO-P) ... Algorithm 2 Bandit-BMO-Zooming (Bandit-BMO-Z)
Open Source Code No The paper does not contain any explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets No The paper uses well-known mathematical functions (Himmelblau’s function and the Styblinski-Tang function) for experiments, which are defined by equations and not typical datasets requiring external access. Therefore, no concrete access information for a dataset is provided.
Dataset Splits No The paper does not provide specific details about training, validation, or test dataset splits.
Hardware Specification No The paper does not provide any specific details about the hardware used for running experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers.
Experiment Setup Yes For the Bandit-BMO-P algorithm, we use ϵ = 0.01, η = 0.001, total number of trials T = 10000. For Bandit-BMO-Z algorithm, we use α = 1, ϵ = 0.01, η = 0.001, number of episodes T = 2500, with four arm trials in each episode.