Multi-Fidelity Multi-Objective Bayesian Optimization: An Output Space Entropy Search Approach

Authors: Syrine Belakaria, Aryan Deshwal, Janardhan Rao Doppa10035-10043

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on several synthetic and real-world benchmark problems show that MF-OSEMO, with both approximations, significantly improves over the state-of-the-art single-fidelity algorithms for multi-objective optimization.
Researcher Affiliation Academia Syrine Belakaria, Aryan Deshwal, Janardhan Rao Doppa School of EECS, Washington State University {syrine.belakaria, aryan.deshwal, jana.doppa}@wsu.edu
Pseudocode Yes Algorithm 1 MF-OESMO Algorithm
Open Source Code No The paper mentions employing code for baselines from the BO library Spearmint, providing a link to its GitHub repository (https://github.com/HIPS/Spearmint/tree/PESM). However, it does not state that the source code for the proposed MF-OSEMO method is open-source or publicly available.
Open Datasets Yes Synthetic benchmarks. We construct two synthetic benchmark problems using a combination of commonly employed benchmark functions for multi-fidelity and single-objective optimization 2, and two of the known general MO benchmarks (Habib, Singh, and Ray 2019). Their complete details are provided in Table 2. Footnote 2: https://www.sfu.ca/ ssurjano/optimization.html. We consider a design space of No C dataset consisting of 1024 implementation of a network-on-chip (Che et al. 2009).
Dataset Splits No The paper does not provide specific training/validation/test dataset splits with percentages, sample counts, or citations to predefined splits. It describes initialization of models and continuous evaluation on benchmark problems rather than a fixed split methodology for model evaluation.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as exact GPU/CPU models, processor types, or memory amounts.
Software Dependencies No The paper mentions using 'MF-GP models', 'squared exponential (SE) kernels', 'Spearmint' for baselines, and 'NSGA-II algorithm'. However, it does not provide specific version numbers for any of these software components or libraries.
Experiment Setup Yes The hyper-parameters are estimated after every 5 function evaluations. We initialize the MF-GP models for all functions by sampling initial points at random from a Sobol grid. We Initialise each of the lower fidelities with 5 points and the highest fidelity with only one point.