genens

genens is an AutoML system for pipeline optimization based on developmental genetic programming.

Installation

Clone the repository.

git clone https://github.com/gabrielasuchopar/genens.git
pip install genens

Using genens

As for now, the GenensClassifier is ready to be used. It has an interface similar to other scikit-learn estimators. When fit() is called, the evolutionary optimization is run. After it finishes, predict() produces a prediction with the best of optimized pipelines. Alternatively, you can call get_best_pipelines() to get pipelines from the pareto front.

from genens import GenensClassifier
from sklearn.datasets import load_iris()

iris = load_iris()
train_X, test_X, train_y, test_y = train_test_split(iris.data, iris.target, test_size=0.25)

clf = GenensClassifier()
clf.fit(train_X, train_y)
... # process of evolution

pred = clf.predict(test_X)

Tests

Directory ./genens/tests contains scripts for running dataset tests and produce data about evolution process along with pickle files of best optimized pipelines. Sample config files are included in ./genens/tests/config.

• Run genEns on a dataset specified in the config file.

python ./genens/tests/run_datasets.py --out OUT_DIR config CONFIG

• Runs genEns on the OpenML-CC18 benchmarking suite

python ./genens/tests/run_openml.py --out OUT_DIR --config CONFIG

More tests are to be included in later releases.