Files
cmdla/Lessons/10-12/UNM.m
2024-07-30 14:43:25 +02:00

291 lines
8.4 KiB
Matlab

function [ x , status ] = UNM( f , varargin )
%function [ x , status ] = UNM( f , x , eps , finf , MaxFeval )
%
% Apply the pure, non-globalised Newton Method for the minimization (or
% maximization, since the method does not distinguish between the two) of
% the provided one-dimensional (Univariate) function f, which must have
% the following interface:
%
% [ v , varargout ] = f( x )
%
% Input:
%
% - x is either a scalar real denoting the input of f(), or [] (empty).
%
% Output:
%
% - v (real, scalar): if x == [] this is the best known lower bound on
% the global optimum of f() on the standard interval in which f() is
% supposed to be minimised (see next). If x ~= [] then v = f(x).
%
% - g (real, either scalar or a [ 2 x 1 ] matrix denoting an interval) is
% the first optional argument. This also depends on x. if x == [] then
% g is a [ 2 x 1 ] matrix denoting the standard interval in which f()
% is supposed to be minimised (into which v is the minimum). f() is
% never called with x ~= [].
%
% - H (real, scalar) is the second optional argument. This must only be
% specified if x ~= [], and it is the second derivative h = f''(x).
% If no such information is available, the function throws error.
%
% IMPORTANT NOTE: the function requires f() to be able to provide both the
% first and the second derivative.
%
% The other [optional] input parameters:
%
% - x: (either a real scalar or [], default []): the starting point; if
% x == [], the left extreme of default range point provided by f() is
% used.
%
% - eps (real scalar, default value 1e-6): the accuracy in the stopping
% criterion: the algorithm is stopped when a point is found such that
% the absolute value of the derivative is less than or equal to eps.
%
% - finf (real scalar, default value 1e+8): since the non-globalised
% Newton Method may diverge, a very rough divergence test is
% implemented whereby if | f( x ) | >= finf then the algorithm is
% stopped with an error condition.
%
% - MaxFeval (integer scalar, default value 30): the maximum number of
% function evaluations (hence, iterations will be not more than
% MaxFeval - 2 because at each iteration one function evaluation is
% performed, except in the first one when two are).
%
% Output:
%
% - x (real scalar): the best solution found so far.
%
% - status (string): a string describing the status of the algorithm at
% termination
%
% = 'optimal': the algorithm terminated having proven that x is a(n
% approximately) optimal solution, i.e., the diameter of the
% restricted range is less than or equal to delta.
%
% = 'stopped': the algorithm terminated having exhausted the maximum
% number of iterations: x is the best solution found so far, but not
% necessarily the optimal one
%
% = 'error': the algorithm found a numerical error that prevents it from
% continuing optimization, such as finding f''( x ) very close to 0,
% or it is found to be diverging (see finf above).
%
%{
=======================================
Author: Antonio Frangioni
Date: 29-09-21
Version 0.20
Copyright Antonio Frangioni
=======================================
%}
Plotg = 3;
% 1 = the function value / gap are plotted
% 2 = the function and the second-order model are plotted
% 3 = the function, the first-order model and the second-order model are
% plotted (the first-order model has no role in the algorithm, but
% this shows how much better the second-order model is than the
% first-order one)
% all the rest: nothing is plotted
Interactive = true; % if we pause at every iteration
% reading and checking input- - - - - - - - - - - - - - - - - - - - - - - -
% - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
if ~ isa( f , 'function_handle' )
error( 'f not a function' );
end
[ fStar , range ] = f( [] );
if isempty( varargin ) || isempty( varargin{ 1 } )
x = range( 1 );
else
x = varargin{ 1 };
if ~ isreal( x ) || ~ isscalar( x )
error( 'x not a real scakar' );
end
end
if length( varargin ) > 1
eps = varargin{ 2 };
if ~ isreal( eps ) || ~ isscalar( eps )
error( 'eps is not a real scalar' );
end
else
eps = 1e-6;
end
if length( varargin ) > 2
finf = varargin{ 3 };
if ~ isreal( finf ) || ~ isscalar( finf )
error( 'finf is not a real scalar' );
end
if finf <= 0
error( 'finf must be in > 0' );
end
else
finf = 1e+8;
end
if length( varargin ) > 3
MaxFeval = round( varargin{ 4 } );
if ~ isscalar( MaxFeval )
error( 'MaxFeval is not an integer scalar' );
end
else
MaxFeval = 30;
end
% initializations - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
% - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
feval = 0;
status = 'optimal';
fbest = Inf;
if Plotg == 1
gap = [];
end
fprintf( 'Univariate Newton''s Method\n');
if fStar > - Inf
fprintf( 'feval\trel gap\t\tx\t\tf(x)\t\tf''(x)\t\tf''''(x)\n');
else
fprintf( 'feval\tfbest\t\tx\t\tf''(x)\t\tf''(x)\t\tf''''(x)\n');
end
% main loop - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
% - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
while true
% compute f( x ), f'( x ), f''( x ) - - - - - - - - - - - - - - - - - -
[ fx , f1x , f2x ] = f( x );
feval = feval + 1;
if fx < fbest
fbest = fx;
end
if abs( f2x ) <= 1e-16
status = 'stopped';
fprintf( 'numerical issue: f''''(x) = %1.4e\n' , f2x );
break;
end
% main logic- - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
xn = x - f1x / f2x;
[ fx , f1x ] = f( x ); % compute f( x ) and f'( x )
% output statistics - - - - - - - - - - - - - - - - - - - - - - - - - -
if fStar > - Inf
gapk = ( fbest - fStar ) / max( [ abs( fStar ) 1 ] );
if Plotg == 1
gap( end + 1 ) = gapk;
semilogy( gap , 'Color' , 'k' , 'LineWidth' , 2 );
xlim( [ 0 35 ] );
ylim( [ 1e-15 inf ] );
ax = gca;
ax.FontSize = 16;
ax.Position = [ 0.03 0.07 0.95 0.92 ];
ax.Toolbar.Visible = 'off';
end
else
gapk = fbest;
end
fprintf( '%4d\t%1.4e\t%1.8e\t%1.4e\t%1.4e\t%1.4e\n' , feval , ...
gapk , x , fx , f1x , f2x );
if Plotg > 1
xm = min( [ x xn ] ) - abs( x - xn ) / 5;
xp = max( [ x xn ] ) + abs( x - xn ) / 5;
warning( 'off' , 'all' );
fplot( @(x) f( x ) , [ xm xp ] , 'Color' , 'k' , 'LineWidth' , 1 );
xlim( [ xm xp ] );
yticks( [] );
ax = gca;
ax.FontSize = 16;
ax.Toolbar.Visible = 'off';
hold on;
if Plotg == 3
% first-order model is
% f( y ) = f( x ) + f'( x )( y - x )
% = [ f( x ) - f'( x ) x ]
% + f'( x ) y
b = f1x;
c = fx - f1x * x;
fplot( @(x) b * x + c , [ xm xp ] , ...
'Color' , 'r' , 'LineWidth' , 1 );
end
% second-order model is
% f( y ) = f( x ) + f'( x )( y - x ) + f''( x )( y - x )^2 / 2
% = [ f( x ) - f'( x ) x + f''( x ) x^2 / 2 ]
% + [ f'( x ) - f''( x ) x ] y
% + f''( x )y^2 / 2
a = f2x / 2;
b = f1x - f2x * x;
c = fx - f1x * x + f2x * x^2 / 2;
fplot( @(x) a * x^2 + b * x + c , [ xm xp ] , ...
'Color' , 'b' , 'LineWidth' , 1 );
if x < xn
xticks( [ xm x xn xp ] );
xticklabels( { num2str( xm , '%1.1g' ) , 'x^k' , ...
'x^{k+1}' , num2str( xp , '%1.1g' ) } );
else
xticks( [ xm xn x xp ] );
xticklabels( { num2str( xm , '%1.1g' ) , 'x^{k+1}' , ...
'x^k' , num2str( xp , '%1.1g' ) } );
end
warning( 'on' , 'all' );
hold off;
end
% stopping criteria - - - - - - - - - - - - - - - - - - - - - - - - - -
if abs( f1x ) <= eps
break;
end
if feval > MaxFeval
status = 'stopped';
break;
end
if abs( fx ) > finf
status = 'error';
break;
end
if Interactive
pause;
end
% iterate - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
x = xn;
end
% end of main loop- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
% - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
% - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
% - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
end % the end- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -