• • Overview
• Links
• • Installation
  • • Installation from PyPI
  • Search Space
  • Quick Start
  • Benchmark
  • Contributing
  • License
  • TODO

XBBO is an an effective, modular, reproducible and flexible black-box optimization (BBO) codebase, which aims to provide a common framework and benchmark for the BBO community.

This project is now supported by PengCheng Lab.

---

Overview | Links |Installation | Quick Start | Benchmark |Contributing | License

For more information and API usages, please refer to our Documentation.

Overview

XBBO decouples the search algorithm from the search space and provides a unified search space interface, allowing developers to focus on the search algorithm.

We provide these black box optimization algorithms as follows:

Search Algorithm	Docs	Official Links	multi-fideility	transfer	multi-obj
Random	Random
Bayesian Optimization	BO
TPE	TPE	hyperopt
BORE	BORE	ltiao/bore
Anneal	Anneal
CEM	CEM
Diffenential Evolution	DE
CMA-ES	CMA-ES	CMA-ES/pycma
NSGA-II					√
Regularized EA	REA	Google-Research
PBT
TuRBO	TuRBO
LaMCTS	LaMCTS	facebookresearch
HyperBand			√
BOHB			√
DEHB			√
MFES-BO			√
TST-R				√
TAF				√
TAF(RGPE)				√
RMoGP				√
RGPE(mean)				√
PSO
XNES
LFBO	LFBO	lfbo-ml/lfbo

Links

• Documentation
• Pypi package

Installation

Python >= 3.7 is required.

Installation from PyPI

To install XBBO from PyPI:

pip install xbbo

For detailed instructions, please refer to Installation.md

Search Space

XBBO uses ConfigSpace as a tool to define search space. Please see ConfigSpace for how to define a search space.

Quick Start

note:XBBO default minimize black box function. All examples can be found in examples/ folder.

Here we take optimizing a quadratic function as a toy example:

from ConfigSpace import ConfigurationSpace
from ConfigSpace.hyperparameters import \
    CategoricalHyperparameter, UniformFloatHyperparameter, UniformIntegerHyperparameter

from xbbo.search_algorithm.bo_optimizer import BO

def custom_black_box_func(config):
    '''
    define black box function:
    y = x^2
    '''
    return config['x'] ** 2

def custom_search_space():
    '''
    define search space
    '''
    configuration_space = ConfigurationSpace()
    configuration_space.add_hyperparameter(UniformFloatHyperparameter('x', -10, 10, default_value=-3))
    return configuration_space

if __name__ == "__main__":
    MAX_CALL = 30

    cs = custom_search_space()

    # specify black box optimizer
    hpopt = BO(space=cs, suggest_limit=MAX_CALL)
    # ---- Begin BO-loop ----
    for i in range(MAX_CALL):
        # suggest
        trial_list = hpopt.suggest() # defalut suggest one trial
        # evaluate 
        obs = custom_black_box_func(trial_list[0].config_dict)
        # observe
        trial_list[0].add_observe_value(obs)
        hpopt.observe(trial_list=trial_list)
        
        print(obs)
    
    print('find best (value, config):{}'.format(hpopt.trials.get_best()))

Please refer to Quick Start.md for more information.

Benchmark

XBBO provides an easy-to-use benchmark tool, users can easily and quickly test the performance of the variety black-box algorithms on each test problem. Clik here for more information.

Contributing

We welcome contributions to the library along with any potential issues or suggestions.

Please refer to Contributing.md in our docs for more information.

License

This project is released under the MIT license.

TODO

• 文档完善
• Logger
• parallel