Due 12/6/16/T in class.
Questions from Text 3rd Ed.

18.4b: Use a parameter alpha between [0,1], to describe an error function returned at leaf.
alpha is such that for =1 error is n (supposed to be all positive), and =0 means error is p (supposed to be all negative). Write such an error function E first, but for Q18.4b, the aplha parts should be squared. Note, p and n are not squared in the equation for E (sum of squared errors, not sqaure of summed errors).
Optimize E for alpha, to find the value of alpha (in terms of p and n) where E becomes min.

18.5. Just show that the information gain is zero for the given condition. Positive-ness for not using the condition need not be answered.

20.5. Note, log-likelihood.