DEMO_SVM |
Demo on Support Vector Machines.
Synopsis:
demo_svm
Description:
DEMO_SVM demonstrates algorithms training the binary
SVM classifier L1-soft and L2-soft margin [Vapnik95]
[Cris00]. The input training vectors must be 2-dimensional
and can be interactively created by the user.
Following algorithms can be tested:
- Sequential Minimal Optimizer (SMO) for L1-norm soft margin.
- QP solver (quadprog) used to train SVM with L2-norm soft margin.
- Kernel Perceptron for separable hyperplane.
Control:
Algorithm - algorithm for testing.
Kernel - non-linear kernel.
Kernel argument - argument of the non-linear kernel.
C-constant - trade-off (regularization) constant.
parameters - parameters of the selected algorithm.
background - if selected then the background color
denotes the sign and the intenzity denotes the value
of the found decision function.
FIG2EPS - exports screen to the PostScript file.
Load data - loads input training sets from file.
Create data - calls program for creating point sets.
Reset - clears the screen.
Train SVM - trains and displays the SVM classifer.
Info - calls the info box.
Close - close the program.
See also
SMO, SVMQUADPROG, KPERCEPTR.
About: Statistical Pattern Recognition Toolbox
(C) 1999-2003, Written by Vojtech Franc and Vaclav Hlavac
Czech Technical University Prague
Faculty of Electrical Engineering
Center for Machine Perception
Modifications:
2-june-2004, VF
18-July-2003, VF
6-march-2002, V.Franc
23-oct-2001, V.Franc