Private Outsourced Bayesian Optimization

Authors: Dmitrii Kharkovskii, Zhongxiang Dai, Bryan Kian Hsiang Low

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

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically evaluate the performance of our PO-GP-UCB algorithm with synthetic and real-world datasets (Section 4). In this section, we empirically evaluate the performance of our PO-GP-UCB algorithm using four datasets including a synthetic GP dataset, a real-world loan applications dataset, a real-world property price dataset and, in Appendix A, the Branin-Hoo benchmark function.
Researcher Affiliation Academia Dmitrii Kharkovskii 1 Zhongxiang Dai 1 Bryan Kian Hsiang Low 1 [...] 1Department of Computer Science, National University of Singapore, Republic of Singapore.
Pseudocode Yes Algorithm 1 PO-GP-UCB (The curator part) and Algorithm 2 PO-GP-UCB (The modeler part)
Open Source Code No The paper does not provide any explicit statements about releasing its source code or links to a code repository for the methodology described.
Open Datasets Yes We use the public data from https://www.lendingclub.com/ and We use the public data from https://www.ura.gov.sg/real Estate IIWeb/transaction/ search.action.
Dataset Splits No The paper describes the total size of the datasets and how they are used in the Bayesian Optimization process, but it does not specify explicit training, validation, or test dataset splits in terms of percentages or sample counts.
Hardware Specification No The paper does not specify any hardware details such as GPU/CPU models, memory, or specific computing environments used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., programming language versions, library versions, or solver versions) needed to replicate the experiment.
Experiment Setup Yes We set the GP-UCB parameter δucb = 0.05 (Theorem 3) and normalize the inputs to have a maximal norm of 25 in all experiments. The GP hyperparameters are learned using maximum likelihood estimation (Rasmussen & Williams, 2006). The function to maximize is sampled from a GP with the GP hyperparameters µx = 0, l = 1.25, σ2 y = 1 and σ2 n = 10 5. We set the parameter r = 10 (Algorithm 1), DP parameter δ = 10 5 (Definition 2) and the GP-UCB parameter T = 50 for this experiment.