High-dimensional Experimental Design and Kernel Bandits

Authors: Romain Camilleri, Kevin Jamieson, Julian Katz-Samuels

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 2.5. Empirical evaluation of allocation methods We briefly describe illustrative experiments (see the supplementary material for more details). G-optimal design experiment: ... Figure 2 depicts the results... G-optimal design in an RKHS: ... Figure 3 depicts the results... RIPS vs. IPS: ... Figure 4 shows that as m grows, the performance of IPS degrades relative to RIPS...
Researcher Affiliation Academia Romain Camilleri 1 Julian Katz-Samuels 2 Kevin Jamieson 1 1Paul G. Allen School of Computer Science and Engineering, University of Washington, Seattle, WA 2University of Wisconsin, Madison, WI.
Pseudocode Yes Algorithm 1 RIPS for Experimental Designs in an RKHS Input: Finite sets X Rd and V H, feature map φ : Rd H, number of samples τ, regularization γ > 0, robust mean estimator bµ : R R ... Algorithm 2 PTR for Experimental Designs in an RKHS ... Algorithm 3 RIPS for Regret Minimization ... Algorithm 4 RIPS for Pure Exploration
Open Source Code No The paper does not include any explicit statement or link providing access to open-source code for the described methodology.
Open Datasets No The paper describes how experimental data was generated ('We generate x1, . . . , xn by sampling xi N(0, Σ)...' and 'We let X = {0, ( 1/m)2, . . . , (m-1/m)2, 1}...') but does not provide concrete access information (link, DOI, formal citation) to a publicly available or open dataset.
Dataset Splits No The paper describes experimental settings like 'G-optimal design experiment' and 'G-optimal design in an RKHS' but does not specify explicit training, validation, or test dataset splits or cross-validation setup.
Hardware Specification No The paper does not provide any specific details about the hardware used for running its experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers used for the experiments.
Experiment Setup No The paper mentions parameters for the experimental setups such as 'd=50', 'm=500', 'ϕ=0.025', and 'γ=0.005' but does not provide details on specific hyperparameters like learning rates, batch sizes, or optimizer settings typically found in machine learning experimental setups.