sksearchspace

Scikit-learn搜索空间配置，提供对scikit-learn估计器的精选搜索空间。

用法

from sksearchspace import SearchSpace
from sklearn.tree import DecisionTreeClassifier

estimator_space = SearchSpace.for_sklearn_estimator(DecisionTreeClassifier, seed=42)
estimator_space.sample()
# {'criterion': 'entropy','min_samples_leaf': 15, 'min_samples_split': 11}

estimator_space.sample()
# {'criterion': 'entropy', 'min_samples_leaf': 12, 'min_samples_split': 4}

sksearchspace使用ConfigSpace进行采样。可以通过属性访问ConfigSpace配置

estimator_space.configuration
# Configuration space object:
# Hyperparameters:
#   criterion, Type: Categorical, Choices: {gini, entropy}, Default: gini
#   min_samples_leaf, Type: UniformInteger, Range: [1, 20], Default: 1
#   min_samples_split, Type: UniformInteger, Range: [2, 20], Default: 2

以下是如何加载JSON文件的示例

with open("search_space.json", "r") as f:
    estimator_space = SearchSpace(f.read())

许可证

版权（c）2020 Thomas J. Fan

根据MIT许可证分发，pytest是免费和开源软件。