Spearmint
Spearmint is a package to perform Bayesian optimization according to the algorithms outlined in the paper: Practical Bayesian Optimization of Machine Learning Algorithms. Jasper Snoek, Hugo Larochelle and Ryan P. Adams. Advances in Neural Information Processing Systems, 2012
Overview
The new code repository for Spearmint is now available on GitHub. This repository has undergone significant improvements and updates to both its algorithms and engineering. It is designed to perform Bayesian optimization for machine learning algorithms, aiming to minimize an objective through iterative adjustments of parameters.
Features
- Automatic experiment running: The code will run experiments in parallel, spawning new experiments as soon as results come in.
- Modularity: The code is designed to be modular, allowing for the swapping out of different 'driver' and 'chooser' modules.
- Supported languages: The code currently supports Python and Matlab as wrapper languages.
- Configuration file: A configuration file is required, specifying the parameters to be tuned and their respective bounds.
Summary
The new code repository for Spearmint on GitHub offers significant improvements and updates to its algorithms and engineering. It provides a convenient way to perform Bayesian optimization for machine learning algorithms, with automatic experiment running and modularity. The installation process requires specific dependencies to be installed, ensuring optimal performance on different operating systems.