.\" DO NOT MODIFY THIS FILE!  It was generated by help2man 1.41.1.
.TH SVM_LEARN "1" "February 2013" "svm_learn of 0.09" "TinySVM"
.SH NAME
svm_learn \- learns SVM models
.SH SYNOPSIS
.B svm_learn
[\fIoptions\fR] \fItraining-file model-file\fR
.SH DESCRIPTION
TinySVM \- tiny SVM package
Copyright \(co 2000\-2002 Taku Kudo, All rights reserved.
.SS "Solver Type:"
.TP
\fB\-l\fR, \fB\-\-solver\-type\fR=\fIINT\fR
select type of solver.
TYPE:  0 \- C\-SVM (default)
.IP
1 \- C\-SVR
2 \- One\-Class\-SVM (experimental)
.SS "Kernel Parameter:"
.TP
\fB\-t\fR, \fB\-\-kernel\-type\fR=\fIINT\fR
select type of kernel function.
TYPE:  0 \- linear      (w * x)  (default)
.TP
1 \- polynomial
(s w * x + r)^d
.TP
2 \- neural
tanh (s w * x + r)
.TP
3 \- RBF
exp (\fB\-s\fR * ||w\-x||^2)
.TP
4 \- ANOVA
(sum_i [exp(\fB\-s\fR * ||w_i\-x_i||^2)])^d
.TP
\fB\-d\fR, \fB\-\-kernel\-degree\fR=\fIINT\fR
set INT for parameter d in polynomial kernel. (default 1)
.TP
\fB\-r\fR, \fB\-\-kernel\-param\-r\fR=\fIFLOAT\fR
set FLOAT for parameter r in polynomial kernel. (default 1)
.TP
\fB\-s\fR, \fB\-\-kernel\-param\-s\fR=\fIFLOAT\fR
set FLOAT for parameter s in polynomial kernel. (default 1)
.SS "Optimization Parameter:"
.TP
\fB\-m\fR, \fB\-\-cache\-size\fR=\fIFLOAT\fR
set FLOAT for cache memory size (MB). (default 40.0)
.TP
\fB\-c\fR, \fB\-\-cost\fR=\fIFLOAT\fR
set FLOAT for cost C of constraints violation,
trade\-off between training error and margin. (default 1.0)
.TP
\fB\-e\fR, \fB\-\-termination\-criterion\fR=\fIFLOAT\fR
set FLOAT for tolerance of termination criterion.
(default 0.001)
.TP
\fB\-H\fR, \fB\-\-shrinking\-size\fR=\fIINT\fR
set INT for number of iterations variable needs to
be optimal before considered for shrinking. (default 100)
.TP
\fB\-p\fR, \fB\-\-shrinking\-eps\fR=\fIFLOAT\fR
set FLOAT for initial threshold value of shrinking process.
(default 2.0)
.TP
\fB\-f\fR, \fB\-\-do\-final\-check\fR=\fIINT\fR
do final optimality check for variables removed
by shrinking. (default 1)
.TP
\fB\-i\fR, \fB\-\-insensitive\-loss\fR=\fIFLOAT\fR
set FLOAT for epsilon in epsilon\-insensitive loss function
used in C\-SVR cost evaluation. (default 0.1)
.SS "Miscellaneous:"
.TP
\fB\-M\fR, \fB\-\-model\fR=\fIFILE\fR
set FILE, FILE.idx for initial condition model file.
.TP
\fB\-I\fR, \fB\-\-sv\-index\fR
write all alpha and gradient to MODEL.idx.
.TP
\fB\-W\fR, \fB\-\-compress\fR
calculate vector w (w * x + b), instead of alpha.
.TP
\fB\-V\fR, \fB\-\-verbose\fR
set verbose mode.
.TP
\fB\-v\fR, \fB\-\-version\fR
show the version of TinySVM and exit.
.TP
\fB\-h\fR, \fB\-\-help\fR
show this help and exit.
.PP
TinySVM \- tiny SVM package
Copyright \(co 2000\-2002 Taku Kudo, All rights reserved.