bayesian-optimizer

Probability & Statistics for AI

This directory contains statistical methods and probabilistic models underpinning modern AI agents, particularly for experimental design and decision making under uncertainty.

Contents

bayesian_optimization.py

A from-scratch implementation of Bayesian Optimization, the engine behind Self-Driving Labs.

• Gaussian Process (GP): Acts as a surrogate model to estimate the outcome of experiments without actually running them.
• Expected Improvement (EI): Acquisition function that balances Exploration (trying high-uncertainty areas) vs Exploitation (trying high-value areas).
• Usage: Used to optimize chemical reactions, hyperparameters, or biological protocols with minimal trials.

Usage

from bayesian_optimization import BayesianOptimizer

# Optimize a function bounded between 0 and 10
opt = BayesianOptimizer(bounds=[(0, 10)])
next_exp = opt.suggest_next_point()