Online update coeffs. for regression with Gaussian noise.
[loglik,K1,K2] = c_reg_gauss(likpar,y,mx,varx)
[loglik,K1,K2] = c_reg_gauss(likpar,y,mx,varx) - takes the noise
variance (likelihood parameter) together with the mean, variance and the
prior mean at the current input and returns the coefficients for the online
update of the GP quadratic regression.
Parameters:
likpar - the noise variance (from net.likpar).
y- observed noisy output.
mx- mean of the GP marginal at the new input.
varx- the variance of the GP marginal at the new input.
loglik- the logarithm of the averaged likelihood function.
K1,K2 - the first and second derivatives of loglik.
If there are more input points, then all outputs K1,K2,loglik will
be matrices [n,1] where n is the numober of observations
(columns) in the inputs.
The likelihood function is
P(x-y,sigma) = exp(-(x-y)^2/(2*sigma^2))sqrt(2*pi);
(see matlab code for details).
ogp, ogptrain, err_mse, c_reg_lapl, c_class_bin, demogp_reg
Copyright (c) Lehel Csató (2001-2004)