Share hyperparameters' values based on similarity of the training sets and models.

deil87 / automl-genetic

Applying genetic programming to AutoML https://travis-ci.org/deil87/automl-genetic.svg?branch=master

12 stars 3 forks source link

Share hyperparameters' values based on similarity of the training sets and models. #160

Open deil87 opened 5 years ago

deil87 commented 5 years ago

Similarity of the training samples (lets say in bagging) could be calculated just by calculation of their diffs (row id could be used)

deil87 commented 5 years ago

Without sharing....if we have small original population and don't have ability to reduce complexity during mutation we can skip(leave it insufficiently explored) some complexity level