Ax: A Platform for Adaptive Experimentation
Sep 1, 2025·
,,,,,,,,,,,,,,,,·
0 min read
Miles Olson
Co-first author
,Elizabeth Santorella
Co-first author
Louis Tiao
Co-first author
,Sait Cakmak
Co-first author
,David Eriksson
Mia Garrard
Sam Daulton
Maximilian Balandat
Eytan Bakshy
Elena Kashtelyan
Zhiyuan Jerry Lin
Sebastian Ament
Bernard Beckerman
Eric Onofrey
Paschal Igusti
Cristian Lara
Benjamin Letham
Cesar Cardoso
Shiyun Sunny Shen
Andy Chenyuan Lin
Matthew Grange
Abstract
Optimizing industry-scale machine learning systems involves
resource-intensive black-box optimization. Adaptive experimentation
substantially improves the sample efficiency of such tasks compared with
naive baselines (such as grid or random search) by utilizing surrogate
models and sequential optimization algorithms. Ax is an open-source
platform for adaptive experimentation. It is highly extensible and
full-featured, and is used at scale at Meta. We discuss Ax’s design,
usage, and performance. Off the shelf, Ax achieves state-of-the-art
performance in a wide range of synthetic and real-world black-box
optimization tasks in machine learning, engineering, and science.
Type
Publication
Proceedings of the 4th International Conference on Automated Machine Learning (AutoML 2025), ABCD Track
Bayesian Optimization
Adaptive Experimentation
AutoML
Hyperparameter Optimization
Open Source
Machine Learning
Authors
Authors
Research Scientist
My name is Louis Tiao, and I graduated from one of Australia’s
top engineering schools with really good grades.
Now, I’m using my knowledge to help up-and-coming tech companies
make it in this competitive world.