Module bayesian 

Gaussian Process surrogates and Bayesian optimization.

Pure-Rust implementation (no external linear algebra):

• GaussianProcess — RBF-kernel GP regression with Cholesky solve; posterior mean + variance for any query point.
• BayesianOptimizer — sequential optimization using Expected Improvement over a GP surrogate.
• Expr — a tiny safe expression evaluator (x[0] * sin(x[1]) + 1) used as the fitness function, mirroring v26’s mc.optimize tool.

The surrogate gives the Dream cycle / Homeostatic loop cheap response-surface models, and mc.optimize replaces grid search with sample-efficient Bayesian exploration.

Structs

BayesianOptimizer

Sequential Bayesian optimization over a box.

Expr

A tiny safe expression evaluator for fitness functions.

GaussianProcess

Gaussian Process regression with an RBF kernel.

OptimizationStep

One step of the Bayesian optimization trace.

Functions

expected_improvement

Expected Improvement acquisition at a query point.

norm_cdf

Numerically safe inverse-normal CDF (Acklam’s algorithm) and CDF (erf).

norm_pdf

Type Aliases

OptimizeResult

Result of a Bayesian optimization run: full trace + best (params, fitness).