PNN

Reference:

Qu Y et al. “Product-based neural networks for user response prediction.” in ICDM 2016

Reference code:

• https://github.com/shenweichen/DeepCTR-Torch/blob/master/deepctr_torch/models/pnn.py

• https://github.com/Atomu2014/product-nets/blob/master/python/models.py

class recbole.model.context_aware_recommender.pnn.InnerProductLayer(num_feature_field, device)

Bases: Module

InnerProduct Layer used in PNN that compute the element-wise product or inner product between feature vectors.

forward(feat_emb)

Parameters:

feat_emb (torch.FloatTensor) – 3D tensor with shape: [batch_size,num_pairs,embedding_size].

Returns:

The inner product of input tensor. shape of [batch_size, num_pairs]

Return type:

inner_product(torch.FloatTensor)

training: bool

class recbole.model.context_aware_recommender.pnn.OuterProductLayer(num_feature_field, embedding_size, device)

Bases: Module

OuterProduct Layer used in PNN. This implementation is adapted from code that the author of the paper published on https://github.com/Atomu2014/product-nets.

forward(feat_emb)

Parameters:

feat_emb (torch.FloatTensor) – 3D tensor with shape: [batch_size,num_pairs,embedding_size].

Returns:

The outer product of input tensor. shape of [batch_size, num_pairs]

Return type:

outer_product(torch.FloatTensor)

training: bool

class recbole.model.context_aware_recommender.pnn.PNN(config, dataset)

Bases: ContextRecommender

PNN calculate inner and outer product of feature embedding. You can choose the product option with the parameter of use_inner and use_outer

calculate_loss(interaction)

Calculate the training loss for a batch data.

Parameters:

interaction (Interaction) – Interaction class of the batch.

Returns:

Training loss, shape: []

Return type:

torch.Tensor

forward(interaction)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

predict(interaction)

Predict the scores between users and items.

Parameters:

interaction (Interaction) – Interaction class of the batch.

Returns:

Predicted scores for given users and items, shape: [batch_size]

Return type:

torch.Tensor

reg_loss()

Calculate the L2 normalization loss of model parameters. Including weight matrices of mlp layers.

Returns:

The L2 Loss tensor. shape of [1,]

Return type:

loss(torch.FloatTensor)

training: bool