Optimization, fast and slow: optimally switching between local and Bayesian optimization

Authors: Mark McLeod, Stephen Roberts, Michael A. Osborne

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We develop the first Bayesian Optimization algorithm, BLOSSOM, which selects between multiple alternative acquisition functions and traditional local optimization at each step. This is combined with a novel stopping condition based on expected regret. This pairing allows us to obtain the best characteristics of both local and Bayesian optimization, making efficient use of function evaluations while yielding superior convergence to the global minimum on a selection of optimization problems, and also halting optimization once a principled and intuitive stopping condition has been fulfilled.
Researcher Affiliation Academia 1Department of Engineering Science, University of Oxford 2Oxford-Man Institute of Quantitative Finance. Correspondence to: Mark Mc Leod <markm@robots.ox.ac.uk>.
Pseudocode Yes Algorithm 1 Positive Definite Test; Algorithm 2 Positive Definite Sphere Radius
Open Source Code No The paper does not provide any specific links or explicit statements about the release of its own source code for the methodology described.
Open Datasets Yes We now give results for several common test objectives for global optimization, illustrated in Figure 5. [...] We use BLOSSSOM to optimize the input and output scale hyperparameters of a Gaussian Process using 6 months of half hourly measurements of UK electricity demand during 2015 1. 1www2.nationalgrid.com/UK/Industryinformation/Electricity-transmission-operational-data/Dataexplorer
Dataset Splits No The paper does not explicitly state specific training, validation, and test dataset splits (e.g., percentages or sample counts) for the benchmark functions or the UK electricity demand data.
Hardware Specification No The paper does not specify any particular hardware used for running the experiments (e.g., GPU models, CPU types, or cloud resources).
Software Dependencies No The paper discusses algorithms and models (e.g., Gaussian Processes, BFGS) but does not provide specific version numbers for any ancillary software dependencies (e.g., programming languages, libraries, or frameworks).
Experiment Setup Yes We have selected a gradient estimate of less that 10 6 as our stopping condition, but any other method could be used. [...] For each algorithm we test multiple values of the stopping criteria, shown in the legend as appropriate.