def rbfbasis(self,l,mu=None):
if not mu:
if not self.xtrain:
mu=self.xtrain
else:
print('DEFINE TRAINING DATA or MU')
return(None)
self.feat=basisfuncs.rbf(l,mu)
def logliklihood(self,l):
liklifeat=basisfuncs.rbf(l,self.xtrain)
R=liklifeat(self.xtrain)
Rinv=np.linalg.inv(R+10e-10*np.eye(R.shape[0]))
if self.mu!=0:
self.mu=np.dot(np.dot(np.ones((1,R.shape[0])),Rinv),self.ytrain)/np.dot(np.dot(np.ones((1,R.shape[0])),Rinv),np.ones((R.shape[0],1)))
self.sigma2=1/float(R.shape[0])*np.dot(np.dot((self.ytrain-self.mu).transpose(),Rinv),(self.ytrain-self.mu))
lnlikli=float(R.shape[0]*np.log(self.sigma2)+np.log(np.linalg.det(R)))
return(lnlikli)
def rbfbasis(self,l,mu=None):
if not mu:
if not self.xtrain:
mu=self.xtrain
else:
print('DEFINE TRAINING DATA or MU')
return(None)
self.feat=basisfuncs.rbf(l,mu)
self.model=np.zeros((len(mu),1)) # initialize model
self.m0=self.model
def rbfbasis(self, l):
mu = self.xtrain
self.feat = basisfuncs.rbf(l, mu)
self.model = np.zeros((len(mu), 1)) # initialize model