SVM1D |
Linear SVM for 1-dimensional input data.
Synopsis:
model = svm1d( data )
model = svm1d( data, options )
Description:
model = svm1d( data ) trains the linear SVM binary
classifier for the 1-dimensional training data.
The optimizer is based on a modification of the
Sequential Minimal Optimizer (SMO) [Platt98].
The trainined classfier is defined as
q(x) = 1 if W*x + b >= 0
= 2 if W*x + b < 0
model = svm1d( data, options ) use to set up control
parameters for the SVM and the SMO algorithm.
Input:
data [struct] Input 1-dimensional binary labeled training data:
.X [1 x num_data] Training numbers.
.y [1 x num_data] Labels (1 or 2).
options [struct] Control parameters:
.C [1x1] SVM regularization constant (default C=inf).
.eps [1x1] Tolerance of KKT-conditions (default eps=0.001).
.tol [1x1] Minimal change of variables (default tol=0.001).
Output:
model [struct] Found SVM model:
.Alpha [nsv x 1] Weights.
.b [1x1] Bias of decision function.
.sv.X [1 x nsv] Support vectors.
.W [1x1] Explicit value of the normal vector (scalar).
.nsv [1x1] Number of Support Vectors.
.kercnt [1x1] Number of kernel evaluations (multiplications
in this 1-d linear case) used by the SMO.
.trnerr [1x1] Training classification error.
.margin [1x1] Margin of found classifier.
.cputime [1x1] Used CPU time in seconds.
.options [struct] Copy of used options.
See also
SMO, SVMCLASS, KFD, KFDQP.
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:
17-may-2004, VF
14-may-2004, VF
15-july-2003, VF