one field has many values, like people's interest, how to resolve this?

[jeremyHW]

No description provided.

[waynesuzq]

Hi, I think I got a similar question. When process movielens data, you convert the label to 0/1:
label = tf.math.greater_equal(label, cutoff)

I wonder what if use the default label 1-5 classes, how is the last layer looks like.
Should we concat the output of FM and DNN, and use another FC to fit the multi-classes?