Probabilisitc Matrix Factorization with Tensorflow

Hello everyone,
Can anyone help me build this recommender collaborative filtering model with TensorFlow.
I thought about making U and V embedding layers but i did not know how to define the loss function with those frobinius norms.
R is the total observations sparse matrix.
I is simple a binary matrix where Iij == 1 if user i did rate item j and 0 otherwise.
This is the loss function:

This is the paper link: Probabilistic Matrix Factorization (

Have you seen class NoBaseClassMovielensModel(tf.keras.Model): in tensorflow recommenders ?

1 Like

Thank you for your answer.
No, this is actually the first time I see it.
Can you please help me with implementing the loss function to modify the NoBaseClassMovielensModel (I’m new with TensorFlow)?
Thanks, in advance.

Sorry but I’m not a consulting. That said, there are a lot of tensorflow courses offered by which will give you a solid foundation in tensorflow.

See Courses - DeepLearning.AI

If this matter is urgent, consider approaching freelancers for this task.