Next:
Contents
Contents
MSVMpack
: a Multi-Class Support Vector Machine package
-
http://www.loria.fr/~lauer/MSVMpack
-
D
OCUMENTATION
OF
V
ERSION 1.5
F. Lauer
Contents
Install and setup
Linux and Mac OS X
Note for Mac OS X users.
Windows
User's guide
Quick start with the web interface
Quick start with the command-line tools
trainmsvm
usage
Infinite training.
M-SVM model type.
Kernel function.
Multiple kernel parameters.
Class-dependent values of
for unbalanced data sets.
Cross validation.
Optimization method.
Data normalization.
Data format.
Automatic testing.
Initialization and saving options.
How
.com
files can help.
Retraining.
Model sparse format.
Log files.
Model saved periodically.
predmsvm
usage
msvmserver
usage
Warning:
Setup of the server.
Web interface
Data files.
Training.
Model files and test.
Using
MSVMpack
from Matlab
Custom kernels.
Developer's guide
make
rules and options
Building
MSVMpack
on Windows
Custom kernels
Using the library
A simple example
Structures
Function reference
Matrix format
File formats
Data file format.
Model file format (
.model
).
Output file format (
.outputs
).
Log file format (
.log
).
Kernel parameter file format.
Optimization features
Parallel implementation
Settings.
Design.
Thread implementation.
Working set selection
M-SVM model of Crammer and Singer (CS)
Other M-SVM models
Kernel cache
Settings.
Design.
Cross validation.
Data format-specific and vectorized kernel functions
Settings.
Vectorization.
Some numerical experiments
License and how to cite
Bibliography
About this document ...
lauer 2014-07-03
for unbalanced data sets.