CoolSearch ❄️

A package for optimizing black-box functions.

• Work in progress

Features

• Grid-search
• Random-search
• Saving previous samples
• Runtime estimation
• Polynomial regression model

Usage

The main entry point is the CoolSearch object, that can be instantiated from

• Class constructor CoolSearch(objective, parameters, ...).
  • This is useful to optimize a function
• Model validation? CoolSearch.model_validate(model, parameters, ...)
  • This is useful if we have a model fulfilling some assumptions.

What is needed

• objective is a function y_1,...y_m = f(x_1,...x_n)
• parameters is a mapping from parameter names to values, ranges or options.

Todo

• smaller range grid search
• Gaussian process model

ideas

• genetic search

note

• (note to self) package should be installable as pip install -e .