Can "genrfeatures" function be used with "fitrnet" function?

In "genrfeatures" function, the "TargetLearner" argument has only three options: linear, SVM and ensemble (bag method). I want to do feature engineering and then use the resulting new features table for generating neural network model. But in the documentation it is suggested to use the corresponding regression learner as the TargetLearner is used.

Is it ok to use any TargetLearner with any other Regression Learner method?

Answer 1

[Rohit]

In general, it is recommended to use “TargetLearner” with its applicable Regression Learner method.

“genrfeatures” assumes that generated features are used to train an interpretable linear regression model which is not there in this case.

But you can still use the features which are returned by “TargetLearner” with any of the 3 types and then pass these features to neural network model(“fitrnet”) and evaluate your results.

Please go through the below documentation link for further reference-

https://www.mathworks.com/help/stats/genrfeatures.html