"SVM classifier visualization"

dkopiec

I'm writing master thesis about pattern classification using SVM and one part of my thesis describes learning SVM. I would like to attach simple example with 2-dimensional input space and couple learning vectors split to to classes.

I wanted to show how SVM will split input space to these classes show chart like one of these:
http://www.pymvpa.org/_images/classifier_comparison_plot.png
or
http://analytics.ijs.si/~blazf/photos/svmDemo.png

in other words i wanted to plot value of y(x; w, b) - and marked positions of learning vectors.

How can I get such plot in Rapidminer?

land

Hi,
switch to the support vector table and choose the plot view. Then select your two dimensions for the axis and select the alpha value as color in the scatter plot Color.

Greetings,
  Sebastian

dkopiec

Thank you, but such plot shows me value of y(x;w,b) only for points that were examples.
I would like to show how y(x;w,b) function looks like (as a heat charts or 3d-surface). I would like to show the surface that separate 2 classes - something like:

http://cmp.felk.cvut.cz/cmp/software/stprtool/examples/svmdemo.gif

land

Hi,
this is not possible in RapidMiner, and it doesn't make much sense at all on real world datasets, because they nearly never have exactly two dimensions...
Just to be curious: What do you need this plot for? Teaching?

Greetings,
  Sebastian