Let's first implement memory$-$based Collaborative Filtering. Complete the function below that takes in a user$-$game matrix user$\_$game$\_$mat as returned by read$\_$data and a zero$-$based user index user$\_$idx. It returns a np$.$ndarray $($or a np$.$matrix$)$ of shape $(1,$ user$\_$game$\_$mat.shape$[1])$ that represents the predicted playtime for each game for the user at user$\_$idx. The algorithm is described in the lecture slide titled "Memory$-$based Approach". We will use Cosine Similarity as the similarity measure between two users based on their raw ratings. You do not need to exclude the user at user$\_$idx when averaging the ratings $$

to compute $$

$,$ since we are interested in recommending games that are not currently being played by the user.