How to determine the amount of features in collaborative filtering

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

Hello,
I am refering to this Video:Collaborative Filtering Algorithm

In the prior video it is explained how to determine the rating, if you know the features of the video (e.g. Romance or Action).
In the linked video, it is stated that usually you do not have these features like and need to come up with the values as illustrated in this picture.

ex1
ex1930×323 79.2 KB

The process in general and the motivation is clear. But if I do not have any information about the movies as stated by Andrew, then how can I know, that I have two features at all (not knowing that they are Action and Romance and what values). How do I know that I do not have 3, 4 or 10 features? He does not explain this and only stipulates, that we do not have information about the features.

2

Hello @M_R2,

I think this learner and this learner have asked very similar questions to yours about how we determine the meaning and number of features in the first place before running the algorithm. Can we really determine the meaning? Do we assign the number by knowing the meaning in the first place? Please check them out when you have time, and leave here any follow-ups that you might have and we can discuss them.

Cheers,
Raymond

1 Like
3

Thank you. Your linked posts and asked questions helped me already :slight_smile:

4

You are welcome @M_R2!