6.7.5.2.1.1.1. statsmodels.sandbox.regression.kernridgeregress_class.GaussProcess.__init__

GaussProcess.__init__(x, y=None, kernel=<function kernel_rbf>, scale=0.5, ridgecoeff=1e-10, **kwds)

Parameters:

x : 2d array (N,K)

data array of explanatory variables, columns represent variables rows represent observations

y : 2d array (N,1) (optional)

endogenous variable that should be fitted or predicted can alternatively be specified as parameter to fit method

kernel : function, default: kernel_rbf

kernel: (x1,x2)->kernel matrix is a function that takes as parameter two column arrays and return the kernel or distance matrix

scale : float (optional)

smoothing parameter for the rbf kernel

ridgecoeff : float (optional)

coefficient that is multiplied with the identity matrix in the ridge regression

Notes

After initialization, kernel matrix is calculated and if y is given as parameter then also the linear regression parameter and the fitted or estimated y values, yest, are calculated. yest is available as an attribute in this case.

Both scale and the ridge coefficient smooth the fitted curve.