Tpot: Generating synthetic data for avoiding overfitting

[Anselmoo]

Question

1. Are any plans started to generate synthetic data during the cycle for avoiding overfitting?
2. Or can tpot be connected to external libraries like: imbalanced-learn

[weixuanfu]

Thank you for your suggestion. But in the near future, we don't have a plan of generating synthetic data in TPOT.

You may use some transformers/estimators in imbalanced-learn with a custom configuration in TPOT if they support scikit-learn API.

[Anselmoo]

Thank you for your suggestion. But in the near future, we don't have a plan of generating synthetic data in TPOT.

You may use some transformers/estimators in imbalanced-learn with a custom configuration in TPOT if they support scikit-learn API.

@weixuanfu is there a possibility to include a scikit learning-curve in this module?

[weixuanfu]

For now, you can easily make a learning-curve with fitted_pipeline_ attribute.