PS Huang et al. “Learning Deep Structured Semantic Models for Web Search using Clickthrough Data” in CIKM 2013.
DSSM respectively expresses user and item as low dimensional vectors with mlp layers, and uses cosine distance to calculate the distance between the two semantic vectors.
Calculate the training loss for a batch data.
interaction (Interaction) – Interaction class of the batch.
Training loss, shape: 
- Return type
Defines the computation performed at every call.
Should be overridden by all subclasses.
Although the recipe for forward pass needs to be defined within this function, one should call the
Moduleinstance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.