https://github.com/mbway/turbo

Bayesian optimisation for global black box function optimisation
https://github.com/mbway/turbo

bayesian-optimization gaussian-processes hyperparameter-optimization optimization

Last synced: about 1 year ago
JSON representation

Bayesian optimisation for global black box function optimisation

• Host: GitHub
• URL: https://github.com/mbway/turbo
• Owner: mbway
• License: gpl-3.0
• Created: 2017-08-27T21:30:25.000Z (almost 9 years ago)
• Default Branch: master
• Last Pushed: 2018-07-02T18:08:38.000Z (almost 8 years ago)
• Last Synced: 2025-03-31T21:27:38.553Z (about 1 year ago)
• Topics: bayesian-optimization, gaussian-processes, hyperparameter-optimization, optimization
• Language: Python
• Size: 520 KB
• Stars: 6
• Watchers: 1
• Forks: 1
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# turbo #

turbo is a modular Bayesian optimisation framework which focuses on gathering and storing the intermediate optimisation steps to give insight into the decision making process.

turbo is capable of producing a wide variety of plots and supports many variations of the basic Bayesian optimisation algorithm.

# Algorithm Features #

- Acquisition Functions

    - Probability of Improvement (PI)

    - Expected Improvement (EI)

    - Upper/Lower Confidence Bound (UCB/LCB)

- Pre-Phase 'naive' selectors

    - Random

    - Latin Hypercube Sampling (LHS)

    - Manual

- Surrogate Models

    - Scikit-Learn Gaussian Process

    - GPy Gaussian Process

- Latent Space

    - Fixed warping (e.g. log-transformed or linear map to `[0,1]` etc)

- Fallback

    - Scheduled random samples ("Harmless" Bayesian Optimisation)

    - de-duplication

- Misc

    - able to use the same storage and plotting functionality with random search or any of the available 'naive' samplers

# Dependencies #

all dependencies can be installed with pip, see `requirements.txt`

# Links

- Black Box Optimisation Benchmarking Procedure: 

- python implementations of many benchmarking functions 

## Some other Bayesian optimisation libraries

- https://github.com/fmfn/BayesianOptimization

- https://github.com/thuijskens/bayesian-optimization

- https://github.com/befelix/SafeOpt

- https://github.com/scikit-optimize/scikit-optimize

- https://hyperopt.github.io/hyperopt/

- https://github.com/automl/RoBO

- https://github.com/resibots/limbo