Example of additional learning using atomization
In this PR, the example of the additional learning impemented. Also, some atomization-related fixes were done.
The experiment consits of 3 steps:
- Initial learning for first 1000 rows of data: auto_model {'roc_auc': 0.834}
Best chain:
-
Additional learning (first 5000 rows) for atomized best model as initial chain: auto_model_from_atomized {'roc_auc': 0.848}
-
Additional learning (first 5000 rows) for non-atomized best model as initial chain: auto_model_from_atomized {'roc_auc': 0.844}
