Hi,

I’m taking collaborative filtering classes and need some clarification regarding notation presented on slide.

I don’t understand what w stands for. Is it “list” of users? And what are these values 5 and 0? Does anyone can explain that?

Thanks in advance

Filip

Hello @Filipdcn , Welcome to the community!

The “w” in machine learning, refers to the weights of the neural network, and is normally a float type, like 0.2. `W`

(uppercase) is a vector, like the example that you just mentioned . Think of w as the intercept (a) of a linear equation `y = ax + b`

.

Hello @Filipdcn,

Another way to look at w is that, the w^{(i)} vector represents a series of attributes of user i, and these attribute values are trainable.

For completeness, b^{(i)} is the bias value for user i, and x^{(j)} represents a series of attributes of movie j. Therefore, each user and each movie has its own set of attributes and are all trainable to best recovering the ratings of all user-movie pairs.

This formulation of problem is different from course 1 and 2 where we have *fixed* feature values labeled as x^{(k)} for sample k, and a trainable w standing for model’s trainable parameter.

Raymond