Logo name
Personal tools

SVM-KA layer block

From Piki

  • Currently0.00/5
Jump to: navigation, search

The SVM-KA block is a Support Vector Machine Kernel Adatron classifer.

SVM-KA layer
Image:SVM-KA layer icon.jpg
Input ports1
Output ports1
DeployableYes
WeightsYes
MemoryNo
Interactive GUINo

Contents

Usage

Support Vector Machines (SVMs) are a powerful type of classifiers that can solve very complex classification problems. They can to that better than any other algorithm currently known to mankind, but it comes at a serious cost: SVMs are one of the most resource demanding adaptive algorithms available as well. They scale badly with data size and evaluation of a trained system is extremely slow as well.

The Synapse SVM component is a variety called a Kernel Adatron, which is a binary classification algorithm. It is reasonably fast as far as SVMs go, but it is limited to solving binary classification problems ( problems that only include two classes).

The component has two ports, one (top) that takes the input signal and the other (bottom) that takes the desired classification. The desired classification must be on ordinal (one feature with values -1 or 1) or nominal (two features with values [1 -1] or [-1 1] form.

Settings

The settings can be modified using the settings browser.


SVM-KA Layer settings


  • (Layout)
  • Inputs: Number of input features.
  • Outputs: Number of output features(must be 1 or 2).
  • (Kernel)
  • Kernel: The choice of kernel function.
  • <*>: Kernel function settings.
  • (SupportVector)
  • Limit: Limits the number of support vectors. Recommended to turn on for larger data sets.

GUI details

The SVM-KA has a standard basic interface:

SVM-KA GUI
Image:SVM-KA GUI short.jpg
Image:SVM-KA GUI long.jpg

See also

This page was last modified 00:22, 15 April 2008.  This page has been accessed 5,175 times.  Disclaimers