Add feature request, Multilabel classification and predict_proba

[agerewines]

System information

• OS version/distro: Windows 10
• .NET Version (eg., dotnet --info): 2.1.602
• ML.NET Version: 1.1.0

Hi everyone, I'm looking for help. I developed some ML models in Python and I'm now trying to import them in ML.NET because my final app is in .NET Core.

I started reading some tutorials and follow step-by-step tutorials from Microsoft. But I can't find out if there is a simple way to implement some features, like:

• multilabel classification: I need to predict a label which can have different values that can be right.
• predict probabilities of classes: in multiclass I can't find a way to implement the predict_proba() of sklearn, where I get in return the probabilities of every class. So that I can show the first two or three most probable classes.

If anyone has some tutorials on how to do this things, or knows if they are planned in the next updates of the library.

Otherwise is there a way to easily implements my python models in a .NET core app?

Thanks, Alessandro

[gvashishtha]

@ImGere have you tried exporting your Python models to ONNX and loading them with an ONNX Transformer? https://docs.microsoft.com/en-us/dotnet/api/microsoft.ml.transforms.onnx.onnxtransformer?view=ml-dotnet

[migratingcub]

Is there an update on if ML.Net will support Multi-label classification out of the box?