Constrained Bayesian optimization with noisy experiments
From MaRDI portal
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.
Recommendations
Cites work
- scientific article; zbMATH DE number 1215244 (Why is no real title available?)
- scientific article; zbMATH DE number 6276166 (Why is no real title available?)
- An informational approach to the global optimization of expensive-to-evaluate functions
- Comparison of Kriging-based algorithms for simulation optimization with heterogeneous noise
- Computational Statistics
- Constrained Bayesian optimization with noisy experiments
- Convergence rates of efficient global optimization algorithms
- Doubly robust policy evaluation and optimization
- Efficient global optimization of expensive black-box functions
- Estimating individualized treatment rules using outcome weighted learning
- Gaussian processes for machine learning.
- Global optimization of expensive-to-evaluate functions: An empirical comparison of two sampling criteria
- Global optimization of stochastic black-box systems via sequential kriging meta-models
- High-dimensional integration: The quasi-Monte Carlo way
- Nonparametric guidance of autoencoder representations using label information
- Parallel Bayesian global optimization of expensive functions
- Policy learning with observational data
- Scrambling Sobol' and Niederreiter-Xing points
- The correlated knowledge gradient for simulation optimization of continuous parameters using Gaussian process regression
- The effective dimension and quasi-Monte Carlo integration
- The no-U-turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo
- Using trajectory data to improve Bayesian optimization for reinforcement learning
Cited in
(18)- Probabilistic bisection with spatial metamodels
- Bayesian optimization in a billion dimensions via random embeddings
- Bayesian Optimization Via Barrier Functions
- Robot joint friction compensation learning enhanced by 6D virtual sensor
- Bayesian optimization approaches for identifying the best genotype from a candidate population
- Constrained stochastic blackbox optimization using a progressive barrier and probabilistic estimates
- Budgeted optimization with constrained experiments
- Learning Enabled Constrained Black-Box Optimization
- Stratified Bayesian optimization
- On parallel policies for ranking and selection problems
- Sequential model based optimization of partially defined functions under unknown constraints
- scientific article; zbMATH DE number 7164723 (Why is no real title available?)
- Applying Bayesian optimization with Gaussian process regression to computational fluid dynamics problems
- The Statistical Filter Approach to Constrained Optimization
- Bayesian Optimization via Exact Penalty
- Bayesian optimisation for constrained problems
- Constrained Bayesian optimization with noisy experiments
- Bayesian optimization for policy search via online-offline experimentation
This page was built for publication: Constrained Bayesian optimization with noisy experiments
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1738149)
