infer Strings Features for BayesPointMachineClassifier

I tried to made a model able to predict a string through a list of number and strings, when I tried to give a object for the mapping I can´t because it only recibe a Vector for the feature.

IClassifierMapping<IList<Vector>, int, IList, string, Vector>

Then I made a dictionary for all the strings in the features, and take the index of each words to pass the value in the vector, but when the model was made the results allways are the same, no matter if I change the parammeters.

If I used the ML.net library with the same data and a Stochastic model, I recibe corrects values for the parammeters.

Do you have some example for strings features to the BayesPointMachineClassifier or is not possible for this kind of models take strigns.

Dec 04 '18 02:12 baf2117

There are no examples of this yet. ML.NET has featurizers that convert strings into Vectors. I agree that it would be useful to have an example of combining an ML.NET featurizer with an Infer.NET classifier.

Dec 05 '18 07:12 tminka

What do you recommend me to use features strings? When you could have and example of strings features or a combine example or ML.Net and Infer.NET?

Dec 07 '18 03:12 baf2117