Problem 3.2

[ddas2]

I seem to be getting a weird graph. The number of iterations required is around 100 when sep=0.2, it then drops down and increases again but remains under 30 till sep=1.6, then it jumps to 70 for sep =1.8 and goes down.... The PLA seems to be working as the lines separate the data for all values of sep. But I cant understand why the number of iterations is showing such random variations.

I have taken average over many iterations for a given sep value. And I still get the same pattern.

[alanericy]

I also have a question about this problem: when we change the sep between two semi-circles do we need to randomly generate the 2000 examples again or just move the original data?

[magdon]

You can randomly generate again.