Created by: rodrigo-arenas
What is this Python project?
This is an AutoML package as an alternative from popular methods inside scikit-learn, such as Grid Search and Randomized Grid Search.
Sklearn-genetic-opt uses evolutionary algorithms to choose the set of hyperparameters that optimizes the cross-validation scores, it can be used for both regression and classification problems with a scikit-learn alike API.
What's the difference between this Python project and similar ones?
- It uses AI for the optimization process, instead of brute force approach like GridSearch.
- It adds several features missing in similar packages, worth to mention:
- Callbacks: Allows to monitor, save the models and stop the training when some of several possible criteria is met, such as the model has run for a long time, a threshold metric was achieved, etc. It even allows the user to create a custom callback.
- Plotting: It was several build-in plotting functionalities to help the user understand the optimization process and take decisions over the models.
- Tensorboard: It can log with just a single line of code all the evaluation metrics to a tensorboard instance to monitor the training.
- MLflow: With one single config class, log all the metrics, models, hyperparameters of each run into a MLflow server.
--
Anyone who agrees with this pull request could submit an Approve review to it.