def initialize(self):
Algorithm.initialize(self)
# to store results
# tridiagonal matrix coefficients
self.alpha = np.zeros(self.maxiter + 1)
self.beta = np.zeros(self.maxiter)
self.vectors = np.zeros((self.maxiter + 1, self.n))
self.w = np.zeros(self.n)
# starting point
self.vectors[0] = np.random.randn(self.n)
self.vectors[0] /= np.sqrt(norm2(self.vectors[0]))
def initialize(self):
Algorithm.initialize(self)
# to store results
# tridiagonal matrix coefficients
self.alpha = np.zeros(self.maxiter + 1)
self.beta = np.zeros(self.maxiter)
self.vectors = np.zeros((self.maxiter + 1, self.n))
self.w = np.zeros(self.n)
# starting point
self.vectors[0] = np.random.randn(self.n)
self.vectors[0] /= np.sqrt(norm2(self.vectors[0]))
def update_gradient_norm(self):
self.last_gradient_norm = copy(self.current_gradient_norm)
self.current_gradient_norm = norm2(self.current_gradient)
def update_beta(self):
self.beta[self.iter_] = np.sqrt(norm2(self.w))
def update_beta(self):
self.beta[self.iter_] = np.sqrt(norm2(self.w))
