Участник:Tatarchuk

Материал из MachineLearning.

(Различия между версиями)
Перейти к: навигация, поиск
(Selective Support Vector Machines)
(Selective Support Vector Machines)
Строка 19: Строка 19:
=== Overview ===
=== Overview ===
-
SelSVMs is library of direct modifications of classical Support Vector Machine provided with controlled selectivity property of feature/kernel selection.
+
SelSVMs are modifications of classical Support Vector Machine provided with controlled selectivity property of feature/kernel selection.
The main features of the algorithms are the following:
The main features of the algorithms are the following:
* work on small training sets with large number of features/kernels;
* work on small training sets with large number of features/kernels;
-
* provide feature selection with controlled selectivity level (numeric parameter);
+
* provide feature selection with controlled selectivity level;
* use Matlab optimization solver and Mosek solver;
* use Matlab optimization solver and Mosek solver;
* support precompiled kernels;
* support precompiled kernels;
* return a posteriori probability estimates;
* return a posteriori probability estimates;
 +
=== How to use ===
=== How to use ===
 +
=== Matlab code ===
=== Matlab code ===
 +
=== Publications ===
=== Publications ===

Версия 12:13, 14 октября 2008

Татарчук Александр Игоревич

Содержание

Selective Support Vector Machines

Description

Authors V. Mottl (vmottl@ya.ru) A. Tatarchuk (aitech@ya.ru) A. Eliseev (andreyel@gmail.com)

Developed at Computing Center of Academy of Sceince, Moscow, Russia

Version 1.0

Date 14.10.2008


Overview

SelSVMs are modifications of classical Support Vector Machine provided with controlled selectivity property of feature/kernel selection.

The main features of the algorithms are the following:

  • work on small training sets with large number of features/kernels;
  • provide feature selection with controlled selectivity level;
  • use Matlab optimization solver and Mosek solver;
  • support precompiled kernels;
  • return a posteriori probability estimates;


How to use

Matlab code

Publications

Личные инструменты