The input hyperparameters for a Classifica​tionDiscri​minant classifier are different from those retrieved from the trained model

1 view (last 30 days)
I am training a discriminant model using fitcdiscr function.
discrMdl = fitcdiscr(Xtrain,Ytrain,'discrimType','linear','ClassNames',{'1','2'},
'Delta', Delta,'Gamma',Gamma);
After training, for double checking, the hypeparameters Gamma and Delta are retrived from
But they are different from those in the function arguments above. How does that happen?
I also note that the descriminant model, which was obtained by training the fitcdiscr with 'OptimizeHyperparameters' configured, is different from the retrained model with the identified hyperparameters (using: bestPoint(discrMdl.HyperparameterOptimizationResults)).
I wonder if the returned model for hyperparamters optimization is already the one trained with the optimal hyparameters, since there may be no need for model retraining.

Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!