File:Gradient descent.png

Summary

File:Gradient descent.svg is a vector version of this file. It should be used in place of this PNG file when not inferior.Category:Vector version available

File:Gradient descent.png → File:Gradient descent.svg

For more information, see Help:SVG.

In other languages
Alemannisch  العربية  беларуская (тарашкевіца)  български  বাংলা  català  нохчийн  čeština  dansk  Deutsch  Ελληνικά  English  British English  Esperanto  español  eesti  euskara  فارسی  suomi  français  Frysk  galego  Alemannisch  עברית  हिन्दी  hrvatski  magyar  հայերեն  Bahasa Indonesia  Ido  italiano  日本語  ქართული  한국어  lietuvių  македонски  മലയാളം  Bahasa Melayu  မြန်မာဘာသာ  norsk bokmål  Plattdüütsch  Nederlands  norsk nynorsk  norsk  occitan  polski  prūsiskan  português  português do Brasil  română  русский  sicilianu  Scots  slovenčina  slovenščina  српски / srpski  svenska  தமிழ்  ไทย  Türkçe  татарча / tatarça  українська  vèneto  Tiếng Việt  中文  中文(中国大陆)  中文(简体)  中文(繁體)  中文(马来西亚)  中文(新加坡)  中文(臺灣)  +/−
New SVG image

Description An illustration of the gradient descent method. I graphed this with Matlab
Date 19 November 2004 (original upload date)
Source Transferred from en.wikipedia to Commons.
Author Olegalexandrov at English Wikipedia
 
This diagram was created with MATLAB.
Category:PNG created with MATLAB#Gradient%20descent.png

Moved from en:Wikipedia, converted to png, and added the matlab source code. Oleg Alexandrov 03:33, 23 June 2007 (UTC)

Licensing

Public domain I, the copyright holder of this work, release this work into the public domain. This applies worldwide.
In some countries this may not be legally possible; if so:
I grant anyone the right to use this work for any purpose, without any conditions, unless such conditions are required by law.
Category:Self-published work#Gradient%20descent.pngCategory:PD-self#Gradient%20descent.png

Source code (MATLAB)

 

% Illustration of gradient descent
function main()

% the ploting window
   figure(1);
   clf; hold on;
   set(gcf, 'color', 'white');
   set(gcf, 'InvertHardCopy', 'off');
   axis equal; axis off;

% the box 
   Lx1=-2; Lx2=2; Ly1=-2; Ly2=2;

% the function whose contours will be plotted
   N=60; h=1/N;
   XX=Lx1:h:Lx2;
   YY=Ly1:h:Ly2;
   [X, Y]=meshgrid(XX, YY);
   f=inline('-((y+1).^4/25+(x-1).^4/10+x.^2+y.^2-1)');
   Z=f(X, Y);

% the contours
   h=0.3; l0=-1; l1=20;
   l0=h*floor(l0/h);
   l1=h*floor(l1/h);
   v=[l0:1.5*h:0 0:h:l1 0.8 0.888];
   [c,h] = contour(X, Y, Z, v, 'b'); 

% graphing settings
   small=0.08;
   small_rad = 0.01;
   thickness=1; arrowsize=0.06; arrow_type=2;
   fontsize=13;
   red = [1, 0, 0];
   white = 0.99*[1, 1, 1];

% initial guess for gradient descent
   x=-0.6498; y=-1.0212;

   % run several iterations of gradient descent
   for i=0:4
      H=text(x-1.5*small, y+small/2, sprintf('x_%d', i));
      set(H, 'fontsize', fontsize, 'color', 0*[1 1 1]);

     % the derivatives in x and in y, the step size
      u=-2/5*(x-1)^3-2*x;
      v=-4/25*(y+1)^3-2*y;
      alpha=0.11;
      
      if i< 4
	 plot([x, x+alpha*u], [y, y+alpha*v]);
	 arrow([x, y], [x, y]+alpha*[u, v], thickness, arrowsize, pi/8, ...
	       arrow_type, [1, 0, 0])
	 x=x+alpha*u; y=y+alpha*v;
      end
      
   end
   
% some dummy text, to expand the saving window a bit
   text(-0.9721, -1.5101, '*', 'color', white);
   text(1.5235,   1.1824, '*', 'color', white);
   
% save to eps
   saveas(gcf, 'Gradient_descent.eps', 'psc2')

function arrow(start, stop, thickness, arrow_size, sharpness, arrow_type, color)

% Function arguments:
% start, stop:  start and end coordinates of arrow, vectors of size 2
% thickness:    thickness of arrow stick
% arrow_size:   the size of the two sides of the angle in this picture ->
% sharpness:    angle between the arrow stick and arrow side, in radians
% arrow_type:   1 for filled arrow, otherwise the arrow will be just two segments
% color:        arrow color, a vector of length three with values in [0, 1]

% convert to complex numbers
   i=sqrt(-1);
   start=start(1)+i*start(2); stop=stop(1)+i*stop(2);
   rotate_angle=exp(i*sharpness);

% points making up the arrow tip (besides the "stop" point)
   point1 = stop - (arrow_size*rotate_angle)*(stop-start)/abs(stop-start);
   point2 = stop - (arrow_size/rotate_angle)*(stop-start)/abs(stop-start);

   if arrow_type==1 % filled arrow

% plot the stick, but not till the end, looks bad
      t=0.5*arrow_size*cos(sharpness)/abs(stop-start); stop1=t*start+(1-t)*stop;
      plot(real([start, stop1]), imag([start, stop1]), 'LineWidth', thickness, 'Color', color);

% fill the arrow
      H=fill(real([stop, point1, point2]), imag([stop, point1, point2]), color);
      set(H, 'EdgeColor', 'none')

   else % two-segment arrow
      plot(real([start, stop]), imag([start, stop]),   'LineWidth', thickness, 'Color', color);
      plot(real([stop, point1]), imag([stop, point1]), 'LineWidth', thickness, 'Color', color);
      plot(real([stop, point2]), imag([stop, point2]), 'LineWidth', thickness, 'Color', color);
   end

function ball(x, y, r, color)
   Theta=0:0.1:2*pi;
   X=r*cos(Theta)+x;
   Y=r*sin(Theta)+y;
   H=fill(X, Y, color);
   set(H, 'EdgeColor', 'none');

Original upload log

The original description page was here. All following user names refer to en.wikipedia.
  • 2004-11-19 02:41 Olegalexandrov 471×492×8 (27067 bytes) An illustration of the gradient descent method. I graphed this with Matlab {{PD}}
Category:Optimization Category:Contour plots Category:Files by User:Oleg Alexandrov from en.wikipedia Category:Gradient descent Category:Images with Matlab source code
Category:Contour plots Category:Files by User:Oleg Alexandrov from en.wikipedia Category:Gradient descent Category:Images with Matlab source code Category:Optimization Category:PD-self Category:PNG created with MATLAB Category:Pages using deprecated source tags Category:Self-published work Category:Vector version available