deephyper.hpo

Hyperparameter optimization and neural architecture search subpackage.

Warning

All optimization algorithms are MAXIMIZING the objective function. If you want to MINIMIZE instead, you should remember to return the negative of your objective.

Classes

ArgMaxEstSelection	Selects solution using a surrogate model and acquisition optimizer.
ArgMaxObsSelection	Selects the best solution based on maximum observed objective(s).
CBO	Centralized Bayesian Optimisation Search.
ExperimentalDesignSearch	Centralized Experimental Design Search.
HpProblem	Class to define an hyperparameter problem.
MPIDistributedBO	Distributed Bayesian Optimization Search using MPI to launch parallel search instances.
RandomSearch	Random search algorithm used as an example for the API to implement new search algorithms.
RegularizedEvolution	Regularized evolution algorithm.
Search	Abstract class which represents a search algorithm.
SolutionSelection	Base class for search solution selection strategies.

gmm	A module for a sampler based on a Gaussian Mixture Model.
utils	Utilities from search algorithms.