Re: Computing Pr(f(x) != g(x))
Hi Bipin,
All methods will give similar probability estimates, when programmed correctly. I prefer exact direct methods because complex stochastic methods tend to be very hard to debug, but the perceptron isn't complex.
I normally code complex algorithms in two different ways in order check the numbers. This time I got sloppy, only implemented a direct method and got spanked for it. It is fair.
You may be able to make your sampling more efficient using recursion which can also be used for relaxing an exact method, but it won't make much of a difference for this assignment even with N=1000.
