Analyzing the top K cross-validated models

[Shoeboxam]

I would like to retain, restore or re-fit the top K cross-validated models, to conduct more in-depth analysis than just taking the model with the best score. I don't see any information in the documentation about this, and I'm not seeing an obvious way to plug the hyperparams from automl.cv_results_['params'][0] back into an object to re-fit. Thank you for any pointers.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs for the next 7 days. Thank you for your contributions.

[anneursula]

HI! Did, by any chance, someone find a way to do this? I am currently trying to do the exact same thing..

[mfeurer]

We do now have an example on how to fit a single model https://automl.github.io/auto-sklearn/master/examples/40_advanced/example_single_configuration.html#sphx-glr-examples-40-advanced-example-single-configuration-py

Is this what you are looking for?

[anneursula]

It could be if I understood correctly. My goal is to analyse all algorithms that ran successfully during 'fit()' (with ensemble_size=0) and refit each of those on my data. I'm thinking I could fit automl, save all configurations from cv_results_ and then fit single models with these configurations?
Or is there an easier way to do this?

[mfeurer]

I cannot think of an easier way to do so.

[eddiebergman]

Documented access to individual model pipelines and should be improved in #709, closing for now.