Can the model from Regression Learner be run with matrices as predictors?

Regression Learner let you export the best prediction model which can then be run with a new input dataset. By default it takes as input a table containing the same predictor and response columns as imported into the app. Can it be modified to run also using images (i.e., matrices) for each of the predictors?

Answer 1

[Akshay Kumar]

I understand that you want to use images(matrices) as predictor variables for your regression function.

However, the documentation mentions that predictor variables can be categorical arrays, cell arrays of character vectors, character arrays, logical vectors, or numeric vectors.

To use images as predictor variables, you would have to extract the relevant features. This is done to remove the redundant features and thus reduce the resources required to describe the image.

For example, the Bag of Features class can be used to extract a bag of features from a given input object.This function will return a double matrix containing the features extracted which can be used as predictor variables in Regression Learner.