How can I set priors in Classification Learner with QDA

I am using Classifiaction Learner App to try out various classifiaction techniques. When I run QDA, I would like to set the priors to the default of 1/K (K groups) before I train. I do not see a way to do that.

Answer 1

[Shanmukha Voggu]

Hi Eugene,

According to my knowledge, In Classification Learner App, it is not possible to set the number of groups in QDA(Quadratic Discriminant Analysis)

For greater flexibility, train a discriminant analysis model using fitcdiscr

As a workaround, QDA can be done using fitdiscr function along with KFold Name-Value Pair to randomly partition the data into k sets.

Refer to this for more information.

[Eugene Davis]

Thanks for the feedback. I have only two groups and that is fixed and compatible with Classification Learner. I want to set the prior probabilities for thoise groups. Now it may still be true that I need to call fitdisc. I would hope that custom priors can be added to the App.