Collective Information, Q&A

1. Roughly, what would you expect the Pearson's value of this graph to be?

> 0.9
about 0
< -0.9

2. The main difference between Amazon's Recommendation System and conventional methods is:

User recommendations are not used
A similarity value is computed between items rather than users
A cosine based metric is used instead of Pearson's

3. A false negative is where:

An item is recommended and shouldn't have been
An item isn't recommended and shouldn't have been
An item isn't recommended and should have been

4. Which of these vectors are equivalent under STI transformation?

(4, 10, 6) and (3, 6, 5)
(1, 10, 1) and (2, 4, 2)
(3, 4, 5) and (2, 3, 5)

5. Trust metrics reduce susceptibility to a Sybil attack because:

It's difficult to create multiple identities
Multiple identities can be easily linked as belonging to one person
Maximal flow through the graph is bottle-necked

6. With Google's PageRank the page with the highest number of links to it will be guaranteed to get the highest page ranking

True
False

7. "Black Sheep" have:

Easy to predict tastes
Very idiosyncratic tastes
Sometimes unusual tastes

8. Model based collaborative filtering systems:

Are faster at answering queries, but are less accurate
Are more accurate, but slower
Use more memory, but are more accurate

9. Having a high level of trust of a key in a web of trust means:

The named person is trustworthy
Keys signed by that key are trustworthy
The key is controlled by the named person

10. Along with Mix Nets, the other fundamental building block for anonymous communications is:

The Dining Cryptographer Ring
The Diving Cartographers Rung
The Digging Cardiologists Rang