Description:

Optunity provides a variety of solvers for hyperparameter tuning problems. The software offers a diverse set of solvers to optimize hyperparameters.

The first major release of Optunity (stable). For documentation, please refer to http://docs.optunity.net. Optunity is compatible with Python 2.7 and above.

The following features are available:

Wide variety of solvers:

• particle swarm optimization
• Nelder-Mead
• grid search
• random search
• Sobol sequences
• CMA-ES (requires DEAP and NumPy)
• TPE (requires Hyperopt and NumPy)

Generic cross-validation functionality:

• support for strata and clusters
• folds are reusable for multiple learning algorithm/solver combinations

Various quality metrics for models (score/loss functions).

Univariate domain constraints on hyperparameters.

Support for parallel objective function evaluations.

Support for structured search spaces.

This release provides Optunity functionality in the following environments: * MATLAB R Octave

Changes to previous version:

The following features have been added:

• new solvers
• tree of Parzen estimators (requires Hyperopt)
• Sobol sequences
• Octave wrapper
• support for structured search spaces, which can be nested
• improved cross-validation routines to return more detailed results
• most Python examples are now available as notebooks

BibTeX Entry: Download

Corresponding Paper BibTeX Entry: Download

Supported Operating Systems: Agnostic

Data Formats: Agnostic

Tags: Optimization, Hyperparameter Tuning

Archive: download here

Comments

No one has posted any comments yet. Perhaps you'd like to be the first?

Leave a comment

You must be logged in to post comments.