Hiroki Naganuma
6.
(a).
Are the examples in S independent? Are they identically distributed?
(b).
What is the distribution of yi given xi?
(c).
Write down the log-likelihood objective.
(d).
This objective should look very similar to another objective for empirical risk minimization with some particular loss function. Which loss function is it?
Mean Square Loss
(e).
In the special case where we consider linear functions for fθ, what are the max- imum likelihood parameters θ (i.e. what do we need to learn)? In this case, the maximum likelihood estimator will be equivalent to the output of a classical regression algorithm. Which one?
Linear Regression
7.
