Perceptron Applet

The applet below demonstrates the perceptron learning rule. 'eta' is learning rate, 'a' vector is the augmented vector representing the discriminant (g(x)) function.

Below in the big pink area you can click with mouse to add samples. two class of samples are supported. left mouse adds a sample from blue set and right mouse button adds a sample from red set.You can press on and drag the mouse to move a sample.

Decision boundary is indicated by a black line when possible (sometimes decision boundary defined by weights lie outside of the visible area.) visible area is a rectangle whose top left corner is the point(0,0) and bottom right corner is the point(1,1) in Cartesian coordinate system

• 'random weights' button assigns random values to 'a' vector.
• 'step' button iterates perceptron algorithm. iterations are made according to batch perceptron rule.
• 'reset' button clears the applet for a new trial
• 'add 10 random points' adds 10 random points on the grid.
• on the right you can see information on the changing values of 'a' and 'Jp' vectors.

Pattern Recognition Term Project 2001-2002 fall.