CS 451 Quiz 25
Low rank matrix factorization, online learning, Map Reduce
* Required
Email address
*
Your email
The predictions for a collaborative filtering problem can be written in matrix form as
*
1 point
X * X'
X * Theta'
Theta * Theta'
Collaborative filtering for movie predictions results in a matrix of size M x U, where M is the number of movies and U is the number of users. The matrix is "low rank" because
*
1 point
The number of features (n) is smaller than both M and U
The matrix consists of mostly zeros
To find related movies, we select
*
1 point
x(i), x(j), such that x(i)  x(j) is small
x(i), x(j), such that x(i)  x(j) is large
theta(i), theta(j), such that theta(i)  theta(j) is small
theta(i), theta(j), such that theta(i)  theta(j) is large
Why is mean normalization a good idea in collaborative filtering?
*
1 point
It makes the algorithm converge faster
It ensures that the matrix to be factored has low rank
It results in reasonable default ratings for are users who have not yet rated anything
Online learning algorithms look at each training example only once
*
1 point
True
False
Online learning algorithm can adapt as user preferences change over time
*
1 point
True
False
What was mentioned in the video as a useful thing to predict for an online algorithm used for product search?
*
1 point
Price
Relevance
Clickthrough rate
Given multiple machines, Map Reduce allows
*
1 point
running minibatch gradient descent, where each machine does a gradient update step
running batch gradient descent, where each machine computes a portion of the gradient update step
Map Reduce can be used both with multiple machines and with a single multicore machine
*
1 point
True
False
To use Map Reduce to train a neural net using multiple computers, some computers perform forward propagation while other computers perform backpropagation, each on a portion of the training set.
*
1 point
True
False
Submit
This content is neither created nor endorsed by Google.
Report Abuse

Terms of Service
Forms