I am using a subset of the Jester Dataset to practice building a collab filtering model using matrix factorization method (as shown in chapter 8). I was mainly interested in seeing the predicted matrix (i.e. the dot product of latent factors). I have put together a brute force workflow to calculate the predicted matrix from the model weights and biases. I am little confused by the result as it does not look anything close to what was expected. I was wondering if anyone had any idea where I might be going wrong.
Here is the link to my notebook - github repo for jester collab model
Thanks in advance!