def update(self):
D2 = matrixmultiply(self.D,self.D)
D3 = matrixmultiply(self.D,D2)
cn = trace(D2-D3)/trace(self.D-D2)
if cn < 0.5:
self.D = ((1.0-2.0*cn)*self.D+(1.0+cn)*D2-D3)/(1.0-cn)
else:
self.D = ((1+cn)*D2-D3)/cn
return
def update(self):
D2 = matrixmultiply(self.D, self.D)
D3 = matrixmultiply(self.D, D2)
cn = trace(D2 - D3) / trace(self.D - D2)
if cn < 0.5:
self.D = ((1.0 - 2.0 * cn) * self.D +
(1.0 + cn) * D2 - D3) / (1.0 - cn)
else:
self.D = ((1 + cn) * D2 - D3) / cn
return
def update(self):
Ne_curr = trace(self.D)
D2 = matrixmultiply(self.D,self.D)
Ne2 = trace(D2)
self.Ne_curr = Ne_curr
# Anders claims this works better; I didn't see a difference
#if abs(2*Ne_curr-Ne2-self.Ne) < abs(Ne2-self.Ne):
if Ne_curr < self.Ne:
self.D = 2*self.D-D2
else:
self.D = D2
return
def update(self):
Ne_curr = trace(self.D)
D2 = matrixmultiply(self.D, self.D)
Ne2 = trace(D2)
self.Ne_curr = Ne_curr
# Anders claims this works better; I didn't see a difference
#if abs(2*Ne_curr-Ne2-self.Ne) < abs(Ne2-self.Ne):
if Ne_curr < self.Ne:
self.D = 2 * self.D - D2
else:
self.D = D2
return
def update(self):
D2 = matrixmultiply(self.D,self.D)
Df = matrixmultiply(D2,4*self.D-3*D2)
trf = trace(Df)
Dp = self.I-self.D
Dp2 = matrixmultiply(Dp,Dp)
Dg = matrixmultiply(D2,Dp2)
trg = trace(Dg)
gamma = (self.Ne-trf)/trg
if gamma > 2:
self.D = 2*self.D-D2
elif gamma < 0:
self.D = D2
else:
self.D = Df-gamma*Dg
return
def update(self):
D2 = matrixmultiply(self.D, self.D)
Df = matrixmultiply(D2, 4 * self.D - 3 * D2)
trf = trace(Df)
Dp = self.I - self.D
Dp2 = matrixmultiply(Dp, Dp)
Dg = matrixmultiply(D2, Dp2)
trg = trace(Dg)
gamma = (self.Ne - trf) / trg
if gamma > 2:
self.D = 2 * self.D - D2
elif gamma < 0:
self.D = D2
else:
self.D = Df - gamma * Dg
return
def initialize(self):
efermi = trace(self.F)/self.N
beta = self.Ne/float(self.N)
alpha = min(self.Ne/(self.emax-efermi),
(self.N-self.Ne)/(efermi-self.emin))/float(self.N)
self.D = alpha*(efermi*self.I-self.F) + beta*self.I
self.Dsumold = 0
return
def initialize(self):
efermi = trace(self.F) / self.N
beta = self.Ne / float(self.N)
alpha = min(self.Ne / (self.emax - efermi),
(self.N - self.Ne) / (efermi - self.emin)) / float(self.N)
self.D = alpha * (efermi * self.I - self.F) + beta * self.I
self.Dsumold = 0
return
def converged(self): return abs(trace(self.D) - self.Ne) < self.tol def init_dmat_solver(Method,**opts):
def get_nel(self,efermi,alpha,beta):
return trace(alpha*(efermi*self.I-self.F)*beta*self.I)
def converged(self):
return abs(trace(self.D) - self.Ne) < self.tol
def converged(self):
self.DS = matrixmultiply(self.D,self.S)
self.Ne_curr = trace(self.DS)
return abs(self.Ne_curr - self.Ne) < self.tol
def converged(self):
return abs(trace(self.D) - self.Ne) < self.tol
def get_nel(self, efermi, alpha, beta):
return trace(alpha * (efermi * self.I - self.F) * beta * self.I)
def converged(self):
self.DS = matrixmultiply(self.D, self.S)
self.Ne_curr = trace(self.DS)
return abs(self.Ne_curr - self.Ne) < self.tol
