Package minfx :: Module base_classes :: Class Min
[hide private]
[frames] | no frames]

Class Min

source code


Instance Methods [hide private]
 
__init__(self)
Base class containing the main minimisation iterative loop algorithm.
source code
 
double_test(self, fk_new, fk, gk)
Default base class function for both function and gradient convergence tests.
source code
 
func_test(self, fk_new, fk, gk=None)
Default base class function for the function convergence test.
source code
 
grad_test(self, fk_new, fk, gk)
Default base class function for the gradient convergence test.
source code
 
hessian_type_and_mod(self, min_options, default_type='Newton', default_mod='GMW')
Hessian type and modification options.
source code
 
minimise(self)
Main minimisation iterative loop algorithm.
source code
 
setup_conv_tests(self)
Default base class for selecting the convergence tests.
source code
 
update(self)
Default base class update function.
source code
Method Details [hide private]

__init__(self)
(Constructor)

source code 

Base class containing the main minimisation iterative loop algorithm.

The algorithm is defined in the minimise function. Also supplied are generic setup, convergence tests, and update functions.

double_test(self, fk_new, fk, gk)

source code 

Default base class function for both function and gradient convergence tests.

Test if the minimum function tolerance between fk and fk+1 has been reached as well as if the minimum gradient tolerance has been reached.

func_test(self, fk_new, fk, gk=None)

source code 

Default base class function for the function convergence test.

Test if the minimum function tolerance between fk and fk+1 has been reached.

grad_test(self, fk_new, fk, gk)

source code 

Default base class function for the gradient convergence test.

Test if the minimum gradient tolerance has been reached. Minimisation will also terminate if the function value difference between fk and fk+1 is zero. This modification is essential for the quasi-Newton techniques.

hessian_type_and_mod(self, min_options, default_type='Newton', default_mod='GMW')

source code 

Hessian type and modification options.

Function for sorting out the minimisation options when either the Hessian type or Hessian modification can be selected.

minimise(self)

source code 

Main minimisation iterative loop algorithm.

This algorithm is designed to be compatible with all iterative minimisers. The outline is:

  • k = 0
  • while 1:
    • New parameter function
    • Convergence tests
    • Update function
    • k = k + 1

update(self)

source code 

Default base class update function.

xk+1 is shifted to xk fk+1 is shifted to fk