Work While They Sleep: Exploiting Evaluation Latency for Fully Bayesian Optimization

Key points

• This objective latency has two consequences during optimization: (i) the objective evaluation dominates execution time, and (ii) sample-efficient algorithms are crucial to accelerate development and avoid wasting resources.1
• In this work, we propose ELF-BO, an algorithm that uses the objective evaluation latency to headstart the computation of the next suggestion, allowing for fully Bayesian optimization without incurring substantial decision-time costs.1
• Across synthetic functions and real-world applications, we show that ELF-BO matches the performance of fully Bayesian methods while only incurring decision latency on par with or better than standard BO.1
• Thus, ELF-BO makes fully Bayesian optimization practical in real-world use cases.1

Sources (1)

• [1]Work While They Sleep: Exploiting Evaluation Latency for Fully Bayesian Optimization
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 06:32 PM

  Black-box optimization problems are ubiquitous across science and engineering, often dealing with expensive objective functions.

  This objective latency has two consequences during optimization: (i) the objective evaluation dominates execution time, and (ii) sample-efficient algorithms are crucial to accelerate development and avoid wasting resources.

Extractive summary: sentences quoted from the sources.