def sigma_init(self):
if hasattr(self,'_sigma_init'):
return self._sigma_init
try:
import pydare
return pydare.dlyap(self.A, self.sigma_states)
except ImportError:
return np.linalg.solve(
np.eye(self.n**2) - np.kron(self.A,self.A), self.sigma_states.ravel())\
.reshape(self.n,self.n)
def sigma_init(self):
if hasattr(self, '_sigma_init'):
return self._sigma_init
try:
import pydare
return pydare.dlyap(self.A, self.sigma_states)
except ImportError:
return np.linalg.solve(
np.eye(self.n**2) - np.kron(self.A,self.A), self.sigma_states.ravel())\
.reshape(self.n,self.n)
def fisher_memory_matrix(W, v, npts = 15, use_dlyap=False):
J = np.zeros([npts, npts])
if use_dlyap:
Cn = dlyap(W, np.eye(W.shape[0]))
else:
Cn = gaussian_covariance_matrix(W)
v = v.reshape([W.shape[0], 1])
Cninv = np.linalg.inv(Cn)
Wmat = np.matrix(W)
for k in range(npts):
for j in range(npts):
v1 = Wmat**(j+1) * v
#print 'v1.shape,',v1.shape
v2 = Cninv * v1
#print 'v2.shape,',v2.shape
v3 = np.transpose(Wmat**(k+1)) * v2
#print 'v3.shape,',v3.shape
v4 = np.transpose(v) * v3
#print 'v4.shape,',v4.shape
J[k, j] = np.array(v4).squeeze()
return J
def process_cov(update, injected_cov):
return pydare.dlyap(update, injected_cov)
def process_cov(update, injected_cov):
return pydare.dlyap(update, injected_cov)
