Instructions for image deblurring optimization "race."

1.
Download from web site
 deblur2_data.mat
 deblur2_race.m

2.
Run deblur2_race.m (using toolbox)
to see how the gradient descent (GD) works for image deblurring.
Ask questions about any parts of this example that are unclear to you!

3.
Edit deblur2_race.m to put in (hopefully faster) optimization methods.
Leave the parameters (regularization, tolerances, etc., unchanged),
including the initial image x^(0).

4.
After you get a method to work, report your # of iterations using
the google doc for class 11 under "homework" on web site.

5.
Try to get even faster convergence (fewer iterations).
Report each successful improvement using that same google doc.

Ground rule:
Each iteration should evaluate the gradient of the cost function only
once.  Other than that you can do as many other operations as needed.
For a more realistic challenge we would use "tic" and "toc" to time
the methods, but with different computers in use that is impractical.