Probability calculation in Q8 and Q10
I correctly approximated the P(f(x) != g(x)) by generating a lot of sample points. However, how do you calculate the probability exactly? Thank you in advance for your help!

Re: Probability calculation in Q8 and Q10
I would be curious on how to calculate  or estimate  the average number of iterations for convergence  as well as the probability  perhaps these are really the same question in disguise. It is a very simple procedure, but there is enough random elements to it so that it leaves the realm of what I am familiar with calculating:
As I understand it ... 1) choose N random points in the unit square 2) choose a random line This gives the target 3) start with w(0) = [0;0] 4) run algorithm where in computing w(n+1) a random selection from the misclassified points is used. (perhaps the random choice here is notrelevant) A pointer to where a similar calculation is made would be fine. Thanks! 
Re: Probability calculation in Q8 and Q10
Quote:

Re: Probability calculation in Q8 and Q10
Quote:
It looks a little similar to the relaxation algorithms for solving partial differential equations. In that case, the theory is well studied. Even without knowing that, you could find the scaling law. How the number of iterations changes, on the average, with the number of points. But once you have the MC program, it is easy enough to change the number of points. 
Re: Probability calculation in Q8 and Q10
Quote:

Re: Probability calculation in Q8 and Q10
Your value of g, assuming it doesn't exactly match f, will be a line with a slope and intercept close, though not exactly that of f. Imagine then, two intersecting lines. The area in between the lines correspond to values in X that g misclassifies. The ratio between that area and that bounded by the region [1,1]x[1,1] (area 4) will be the probability of error. Alas, calculating the area between the lines is ugly. The problem would have been much simpler if the domain for X had been points inside the circle at (0,0) with radius 1.

Re: Probability calculation in Q8 and Q10
Quote:
If the input space is 2D, I can see that there are two triangles formed by f(x) and g(x), and the total area of the two triangles include all the errors. So I thought the P(f(x) != g(x)) = area of error / total area of input space...no? Thanks in advance. 
Re: Probability calculation in Q8 and Q10
Quote:

Re: Probability calculation in Q8 and Q10
Argh, I see.
Missed the boundaries stated in the problem. My bad. Thanks. 
All times are GMT 7. The time now is 09:28 AM. 
Powered by vBulletin® Version 3.8.3
Copyright ©2000  2020, Jelsoft Enterprises Ltd.
The contents of this forum are to be used ONLY by readers of the Learning From Data book by Yaser S. AbuMostafa, Malik MagdonIsmail, and HsuanTien Lin, and participants in the Learning From Data MOOC by Yaser S. AbuMostafa. No part of these contents is to be communicated or made accessible to ANY other person or entity.