View Single Post
#2
02-13-2013, 10:00 PM
 vikasatkin Caltech Join Date: Sep 2011 Posts: 39
Discussion of Lecture 11 "Overfitting"

Question: (Slide 11/23) How did you generate the polynomials? How did you choose the coefficients?

Answer: Here is the technical description of the process of generating the target function and the dataset (which may be useful, if you want to reproduce the pictures from the slide 13/23). It was actually described in the "Learning From Data" book on p.123 (section 4.1.2 "Catalysts for Overfitting").

The process of generating the target function depends on two parameters: (degree of the generated polynomials) and (noise level). Of course, you also need --- amount of points in the dataset.

1. Take Legendre polynomials . Note, that they are normalized according to their value at (i.e. ), not their average square.
2. Choose coefficients independently according to the standard normal distribution.
3. Generate points (pick them randomly from , independently from each other).
4. For every point generate the noise .

The target is given by . Here is a normalization constant, which depends only on . It is chosen in such a way, that the mean square value of is equal to 1 (mean with respect to both and choices we made during this process: ). On can compute, that