Dot product for content-based filtering

Course Q&A Machine Learning Specialization Unsupervised Learning, Recommenders, Reinforcement
1

For the content-based filtering, we use dot product for prediction. How if the paired items on the 2 vectors are not relevant, the dot product makes no sense?

image
image1178×480 59 KB

2

If the paired item is not relevant, then its weight value will be near to zero. So the dot product is still a valid operation.

3

if it’s by random, more likely, the pair would be irrelevant, unless there’s mechanism in place to list the features relevant for the pair.

4

A dot product is the classic method of computing similarity between two vectors.

1 Like
5

so the dot product would reshuffle the vectors?

6

No, it just computes the similarity.