Drawing boundary line for PLA

[dsvav]

Hello,

This is not exactly related to HW(which I have submitted).

I want to visually observe how boundary line changes after each iteration i.e I want to know how to draw a boundary line after each iteration of PLA.

Initial or actual boundary line (rep target function) is known to me since it is selected as two random points and I can draw the line. One iteration gives me a new weight vector , so how do I relate this to boundary line.

Thanks and I apologize for any wrong use terminologies.

[rainbow]

Since we use the sign function, I presume the decision boundary is
g(x) = w'x = 0 (including the bias w_0)

One strategy could be to generate values on the grid [-1,1] x [-1,1], and plot where g evaluates to 0. But that is costly for each iteration. Hopefully there is a better approach.

[JohnH]

The following are the equations for the plot of the learned boundary line:



If or is 0, then the corresponding component of the input vector does not contribute to the result and the boundary is a horizontal or vertical line, respectively.

[dsvav]

Thanks JohnH and rainbow.