# Constrained Bayesian Optimization with Noisy Experiments

## TL;DR

This paper introduces a new Bayesian optimization method tailored for noisy, constrained experiments, improving efficiency and accuracy in real-world applications like ranking systems and compiler optimization.

## Contribution

It derives a new expected improvement expression for noisy, constrained Bayesian optimization and develops an efficient approximation method for practical use.

## Key findings

- Outperforms existing methods on synthetic noisy, constrained problems.
- Effectively applied to real-world Facebook experiments.
- Enhances optimization performance in high-noise, real-world scenarios.

## Abstract

Randomized experiments are the gold standard for evaluating the effects of changes to real-world systems. Data in these tests may be difficult to collect and outcomes may have high variance, resulting in potentially large measurement error. Bayesian optimization is a promising technique for efficiently optimizing multiple continuous parameters, but existing approaches degrade in performance when the noise level is high, limiting its applicability to many randomized experiments. We derive an expression for expected improvement under greedy batch optimization with noisy observations and noisy constraints, and develop a quasi-Monte Carlo approximation that allows it to be efficiently optimized. Simulations with synthetic functions show that optimization performance on noisy, constrained problems outperforms existing methods. We further demonstrate the effectiveness of the method with two real-world experiments conducted at Facebook: optimizing a ranking system, and optimizing server compiler flags.