Bayesian
Optimization.
What to do when a grid of experiments is too expensive to run. How the algorithm picks the next trial, and how to drive it from Python.
- Part 1
Understanding Bayesian Optimization Start here
How an algorithm learns from the runs you have already done, and picks the next experiment worth doing.
- Part 2
Testing Bayesian Optimization for Lab Experiments
Bayesian optimization against random sampling against fractional factorial design on a real dataset.
- Part 3
A Technical Guide to Bayesian Optimization
What sits under the hood: surrogate models stand in for your expensive runs, acquisition functions decide where to look next.
- Part 4
Getting Started with Bayesian Optimization (BayBE) Python
A walkthrough of the BayBE package: set up the search space, run the loop, feed each result back in.
- Part 5
Getting Started with Bayesian Optimization (NEXTorch) Python
A walkthrough of the NEXTorch package, from the search space to the next experiment it suggests.
- Part 6
Getting Started with Bayesian Optimization (BoFire) Python
A walkthrough of the BoFire package: define the domain and its constraints, then let it propose the next runs.
- Part 7
How Many Experiments Do I Really Need?
Active learning, your own judgement, or random picks. Which one gets to a good model in the fewest experiments?