MATLAB fct predint for Gaussian-Process-Model

Hi

I'm using the matlab toolbox model-based calibration and I'm trying to fit local models for a Diesel engine. The default model fits inlcude quadratic and cubic polynomials, a hybrid fit with a linear polynomial and a radial-basis function (RBF), and a gaussian process model with a constant (GPM).

After exporting my best model to MATLAB, I try to use the function predint, which is a method for the class mbcPointByPointModel. First I found a bug in MATLAB, now it works fine for polynomials and RBF, but doesn't work for GPM.

Does anyone have an idea why?

code example

%thisModel is an instant of the class mbcPointByPointModel %Xpred are all inputs to the model intervals = predint(thisModel,Xpred)

Answer 1

[Sindhuja Parimalarangan]

This could possibly be because predint cannot compute prediction intervals for non-parametric regression methods and Gaussian Process Regression model in MATLAB is a nonparametric kernel-based probabilistic model.