abstract_process.py 文件源码

def fit(self, X, y):
        """Fit the parameters of the probabilistic process based on the
        available training data.
        """
        # Store the training data (both the inputs and the targets).
        self.X, self.y = X, y
        # Compute the covariance matrix of the observed inputs.
        K = self.kernel.cov(self.X)
        # For a numerically stable algorithm, we use Cholesky decomposition.
        self.L = spla.cholesky(K, lower=True)
        self.alpha = spla.cho_solve((self.L, True), self.y).ravel()
        L_inv = spla.solve_triangular(self.L.T, np.eye(self.L.shape[0]))
        self.K_inv = L_inv.dot(L_inv.T)
        self.beta = self.y.dot(self.alpha)