Modify IDataView in AutoML Experiment After Transform and Before Evaluate

[superichmann]

Is your feature request related to a problem? Please describe. No

Describe the solution you'd like Add the option to modify the idataview (such as in preFeaturizer) but a "postFeaturizer" which will transform the idataview after the Transform has occurred on it inside the experiment and before the evaluation metrics are calculated.

Describe alternatives you've considered Create my own trial runner, if possible and this feature is not planned in automl please provide me with a start code :]

Additional context Some use cases require alteration of the idataview based on the Score column which is not present before Transform is called. Another solution would be to add the possibility to call a custom evaluate function based on LINQ

[superichmann]

?

[luisquintanilla]

Hi @superichmann,

Thanks for this suggestion. At this moment, we don't have plans to add this feature, though as you suggested, creating your own trial runner would be what's recommended here.

Here's some documentation on it.

https://learn.microsoft.com/dotnet/machine-learning/how-to-guides/how-to-use-the-automl-api#create-your-own-trial-runner