SVM applets
Remember that these applets are experimental pedagogic works, bugs are still to be found, features are still to be added. If you have anything constructive to say about them, please send an e-mail to Guillaume Caron.
This applet is a SVM visualization applet. A two-dimension plane is given to you and positive/negative examples are entered by clicking on the mouse (left click=positive, right click=negative) or by using five different presets given to you. Remember presets are additive (select many presets without clearing the plane to see what I mean) You are also given fourkernel functions to try out. You then press the Compute SVM button and the entire plane is classified. The visualization itself is a RGB encoding of this classification.
This simple applet displays a linear SVM in a 2D plane, dynamically updated given examples entered on this plane by the mouse (left click=positive, right click=negative).You can observe the behaviour of the SVM according to an error in the training set for example.