Source Code for Module minimise.newton

 1  from LinearAlgebra import inverse 
 2  from Numeric import copy, dot, matrixmultiply 
 3   
 4  from generic_line_search import generic_line_search 
 5  from generic_minimise import generic_minimise 
 6   
 7   
 8 -class newton(generic_line_search, generic_minimise): 
 9 -        def __init__(self, func, dfunc=None, d2func=None, args=(), x0=None, line_search_algor=None, func_tol=1e-5, maxiter=1000, full_output=0, print_flag=0, a0=1.0, mu=0.0001, eta=0.9): 
10                  "Class for Newton minimisation specific functions." 
11   
12                  self.func = func 
13                  self.dfunc = dfunc 
14                  self.d2func = d2func 
15                  self.args = args 
16                  self.xk = x0 
17                  self.func_tol = func_tol 
18                  self.maxiter = maxiter 
19                  self.full_output = full_output 
20                  self.print_flag = print_flag 
21   
22                  if not line_search_algor: 
23                          raise NameError, "No line search algorithm has been supplied." 
24                  else: 
25                          self.line_search_algor = line_search_algor 
26   
27                  # Set a0. 
28                  self.a0 = a0 
29   
30                  # Line search constants for the Wolfe conditions. 
31                  self.mu = mu 
32                  self.eta = eta 
33   
34                  # Initialise the function, gradient, and hessian evaluation counters. 
35                  self.f_count = 0 
36                  self.g_count = 0 
37                  self.h_count = 0 
38   
39                  # Initialise the warning string. 
40                  self.warning = None 
41   
42                  # The initial Newton function value, gradient vector, and hessian matrix. 
43                  self.fk, self.f_count = apply(self.func, (self.xk,)+self.args), self.f_count + 1 
44                  self.dfk, self.g_count = apply(self.dfunc, (self.xk,)+self.args), self.g_count + 1 
45                  self.d2fk, self.h_count = apply(self.d2func, (self.xk,)+self.args), self.h_count + 1 
46   
47                  # Minimisation. 
48                  self.minimise = self.generic_minimise 
49   
50   
51 -        def backup_current_data(self): 
52                  "Function to backup the current data into fk_last." 
53   
54                  self.fk_last = self.fk 
55   
56   
57 -        def dir(self): 
58                  "Calculate the Newton direction." 
59   
60                  self.pk = -matrixmultiply(inverse(self.d2fk), self.dfk) 
61                  if dot(self.dfk, self.pk) >= 0.0: 
62                          self.pk = -self.dfk 
63   
64   
65 -        def update_data(self): 
66                  "Function to update the function value, gradient vector, and hessian matrix" 
67   
68                  self.xk = copy.deepcopy(self.xk_new) 
69                  self.fk, self.f_count = apply(self.func, (self.xk,)+self.args), self.f_count + 1 
70                  self.dfk, self.g_count = apply(self.dfunc, (self.xk,)+self.args), self.g_count + 1 
71                  self.d2fk, self.h_count = apply(self.d2func, (self.xk,)+self.args), self.h_count + 1 
72