import numpy as np
class GaussianDiscriminant:
def __init__(self,k=2,d=8,priors=None,shared_cov=False):
self.mean = np.zeros((k,d)) # mean
self.shared_cov = shared_cov # using class-independent covariance or not
if self.shared_cov:
self.S = np.zeros((d,d)) # class-independent covariance
else:
self.S = np.zeros((k,d,d)) # class-dependent covariance
if priors is not None:
self.p = priors
else:
self.p = [1.0/k for i in range(k)] # assume equal priors if not given
self.k = k
self.d = d
def fit(self, Xtrain, ytrain):
# compute the mean for each class
if self.shared_cov:
# compute the class-independent covariance
pass # placeholder
else:
# compute the class-dependent covariance
pass
def predict(self, Xtest):
# predict function to get predictions on test set
predicted_class = np.ones(Xtest.shape[0]) # placeholder
for i in np.arange(Xtest.shape[0]): # for each test set example
# calculate the value of discriminant function for each class
for c in np.arange(self.k):
if self.shared_cov:
pass # placeholder
else:
pass
# determine the predicted class based on the values of discriminant function
return predicted_class
def params(self):
if self.shared_cov:
return self.mean[0], self.mean[1], self.S
else:
return self.mean[0],self.mean[1],self.S[0,:,:],self.S[1,:,:]
class GaussianDiscriminant_Diagonal:
def __init__(self,k=2,d=8,priors=None):
self.mean = np.zeros((k,d)) # mean
self.S = np.zeros((d,)) # variance
if priors is not None:
self.p = priors
else:
self.p = [1.0/k for i in range(k)] # assume equal priors if not given
self.k = k
self.d = d
def fit(self, Xtrain, ytrain):
# compute the mean for each class
# compute the variance of different features
pass # placeholder
def predict(self, Xtest):
# predict function to get prediction for test set
predicted_class = np.ones(Xtest.shape[0]) # placeholder
for i in np.arange(Xtest.shape[0]): # for each test set example
# calculate the value of discriminant function for each class
for c in np.arange(self.k):
pass
# determine the predicted class based on the values of discriminant function
return predicted_class
def params(self):
return self.mean[0], self.mean[1], self.S