Nonparametric Bayes Action Set
Kernel
The LA inference of the gpClass action uses either of two kernel types for the Gaussian process in Gaussian process classification: Gaussian or linear. The kernel type is specified by the kernel parameter.
The Gaussian kernel type represents a covariance matrix that is generated by a squared exponential function with specified input data, as shown in the following equation, where 
is the Gaussian kernel bandwidth parameter specified by the sigma subparameter. The Gaussian kernel is versatile enough to separate two classes that are not linearly separable.
The linear kernel type represents a covariance matrix that is generated using a linear function, as shown in the following equation, where c is the constant in the linear function that is specified by the constant subparameter. When the classes in the data are linearly separable, the linear type can yield good results and a faster processing time than the Gaussian type.
