Question 10 states:

"Here, we compute the expected utility of a given item, using the Hacker News formula."

In fact to pass the asserts you need to use the revised formula, which is shown under Q4.

Minor issue,

At Point 2 we're trying to load a matrix but the file doesn't exist.

R_ = np.genfromtxt('../data/interim/ratings_matrix.csv', delimiter=',')

created a boolean array of votes >0 (shape = (671, 9066))
then sum all rows (per column) (shape = (9066, ))

Shouldn't return one dimension array (9066, ) instead of two dimension (1, 9066)?
even after reshaping to (1, 9066) all asserts fail...

On Q12 you need to return the top N in order [largest, 2nd largest, 3rd largest, ...], otherwise the assert won't pass.

If you've done the slicing in a way that changes that, try typing them out in order (you know which is the biggest one from Q11) and see if it passes the assert.

This appears more than once in the learning notebooks:

ndims = R_.ndim
nrows = R_.shape[0]
ncols = R_.shape[0] 
dtype = R_.dtype

Actualy should be

ndims = R_.ndim
nrows = R_.shape[0]
ncols = R_.shape[1]  # one 
dtype = R_.dtype

:)

Hi, in the line:
To clarify, IAna={Bananas,Milk} and (Ana,Bananas)∉R′ . Right?

I guess what is meant is that:
(Ana, Water)∉R′
as the rating for column 1 - Water is missing.