Last active
January 26, 2021 10:16
-
-
Save ShuaiGuo16/932e0288cfc6903b4d81857773d1255a to your computer and use it in GitHub Desktop.
Calculate negative log-likelihood
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
def Neglikelihood(self, theta): | |
"""Negative likelihood function | |
Input | |
----- | |
theta: array, logarithm of the correlation legnths for different dimensions | |
Output | |
------ | |
LnLike: likelihood value""" | |
theta = 10**theta # Correlation length | |
n = self.X.shape[0] # Number of training instances | |
one = np.ones((n,1)) # Vector of ones | |
# Construct correlation matrix | |
K = self.Corr(self.X, self.X, theta) + np.eye(n)*1e-10 | |
inv_K = np.linalg.inv(K) # Inverse of correlation matrix | |
# Mean estimation | |
mu = (one.T @ inv_K @ self.y)/ (one.T @ inv_K @ one) | |
# Variance estimation | |
SigmaSqr = (self.y-mu*one).T @ inv_K @ (self.y-mu*one) / n | |
# Compute log-likelihood | |
DetK = np.linalg.det(K) | |
LnLike = -(n/2)*np.log(SigmaSqr) - 0.5*np.log(DetK) | |
# Update attributes | |
self.K, self.inv_K , self.mu, self.SigmaSqr = K, inv_K, mu, SigmaSqr | |
return -LnLike.flatten() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment