Applets & Web Software

The SVM Applet

Support Vector Machines are learning machines that can perform binary classification (pattern recognition) and real valued function approximation (regression estimation) tasks. Support Vector Machines non-linearly map their n-dimensional input space into a high dimensional feature space. In this high dimesional feature space a linear classifier is constructed.

The CLRC has implemented a Java Applet which demonstrates the potential of the SVM approach. The Applet can be downloaded from this page, or if you just want to see what it can do you can use the links below to run the programme for:

• A pattern recognition classification problem, or;
• A regression estimation problem.

News

A new technique for "hedging" predictions was presented and discussed recently by Alexander Gammerman and Vladimir Vovk at a special meeting of the British Computer Society. The method can be applied to many algorithms, including Support Vector Machines, Kernel Ridge Regression, Kernel Nearest Neighbours and other state-of-the-art methods.

The hedged predictions include confidence measures that are provably valid and it becomes possible to control the number of errors by selecting a suitable confidence level.

The discussants of the technique included Vladimir Vapnik, Alexey Chervonenkis, Glenn Shafer, Zhiyuan Luo and many others.

The paper and the discussion can be found here.

Our Applet is sponsored by

The SVM website is implemented by

SVM Applet Feedback

"Nice visualisation technique."
Dr Michael Thess, Prudential Systems Software GmbH

"This is a great demo...I was trying to compare it with other classification techniques such as Discriminant Analysis, and it does show what SVs can do really well."
Aravind Ganapathiraju, Mississippi State University

"Congratulations! Your applet could be useful for teaching purposes."
Giorgio Valentini, DISI Genova

"Keep up the great work in this important new area! Congratulations."
A. Richard Newton, Dean and the Roy W. Carson Professor in Engineering,
University of California, Berkeley.

"I'm having fun seeing what happens as the various parameters are varied...quite a show..."
Grace Wahba, University of Wisconsin-Madison

The Applet has been very successful. It has been downloaded to date by around 700 individuals affiliated with more than 400 separate companies and institutions in approximately 50 countries across the world (you can see a list of these companies and institutions here) and accessed by many more. The feedback we have received on the performance of the Applet has been universally positive.