Emile Contal - Homepage

doc: gp_downdate

Posterior $\mu(x_i)$ and $\sigma^2(x_i)$ given $X_t \setminus \{x_i\}$

Syntax

 [mu, sigma2] = gp_downdate(Ktt, Yt, i, BayesInv)
 [mu, sigma2] = gp_downdate(Ktt, Yt, i, BayesInv, Ht)

Arguments

Ktt kernel matrix (nt, nt) between the points of Xt
Yt vector (nt, 1) of observations
i indice of removed observation
BayesInv struct array returned by gp_inf(Ht, Ktt, Yt, noise)
Ht matrix (nt, b) of basis data as returned by basis_cst(Xt)

Outputs

mu scalar (1,1) posterior mean $E[f(x_i) \mid X_t\setminus x_i, Y_t \setminus y_i]$
sigma2 scalar (1,1) posterior variance $V[f(x_i) \mid X_t\setminus x_i, Y_t \setminus y_i]$

See also

gp_pred | gp_loolik

web analytics