Rules of engagement:

The data in directory cases/charity/sub_scr is exactly like the data
for Homework 2 except that the targets num/471.dat and num/472.dat
are missing.

The data used for Homework 2 comes from a vendor competition.  What is
in directory cases/charity/sub_big is oversampled from the model set
that was given to the competitors to train their tools.  The data in
directory cases/charity/sub_scr is oversampled from the score set that
was used to judge the competitors.

Build your model for Homework 2 and use it to score the data in 
cases/charity/sub_scr.

I will do the same.

I have the target and will compute my MSE and the MSE of all teams that
turn in a predict.dat file for these data.

If any team beats me, we all go out for beer at the Washington Duke at
my expense.