Understanding fitrgp Output: Difference Between Sigma, SigmaL, and SigmaF in MATLAB?

In MATLAB fitrgp, the displayed hyperparameter values depend on the chosen kernel covariance function:

• Sigma (Noise Std Dev): Represents observation noise standard deviation σ_n.
• SigmaL (Length Scale): Vector of length N (number of predictors) representing the characteristic length scale per predictor feature.
• SigmaF (Signal Std Dev): Scalar factor representing overall process variance output σ_f.

Accessing Trained Hyperparameters:

gprMdl = fitrgp(X, y, 'KernelFunction', 'squaredexponential');
sigma_noise = gprMdl.Sigma;
kernel_params = gprMdl.KernelInformation.KernelParameters;
disp(kernel_params);