hopwise.model.knowledge_aware_recommender.kgat¶

Reference:: Xiang Wang et al. “KGAT: Knowledge Graph Attention Network for Recommendation.” in SIGKDD 2019.
Reference code:: https://github.com/xiangwang1223/knowledge_graph_attention_network

Classes¶

`Aggregator`	GNN Aggregator layer
`KGAT`	KGAT is a knowledge-based recommendation model. It combines knowledge graph and the user-item interaction

Module Contents¶

class hopwise.model.knowledge_aware_recommender.kgat.Aggregator(input_dim, output_dim, dropout, aggregator_type)[source]¶

Bases: torch.nn.Module

GNN Aggregator layer

input_dim¶

output_dim¶

dropout¶

aggregator_type¶

message_dropout¶

activation¶

forward(norm_matrix, ego_embeddings)[source]¶

class hopwise.model.knowledge_aware_recommender.kgat.KGAT(config, dataset)[source]¶

Bases: hopwise.model.abstract_recommender.KnowledgeRecommender

KGAT is a knowledge-based recommendation model. It combines knowledge graph and the user-item interaction graph to a new graph called collaborative knowledge graph (CKG). This model learns the representations of users and items by exploiting the structure of CKG. It adopts a GNN-based architecture and define the attention on the CKG.

input_type¶

ckg¶

all_hs = None¶

all_ts = None¶

all_rs = None¶

matrix_size¶

embedding_size¶

kg_embedding_size¶

layers¶

aggregator_type¶

mess_dropout¶

reg_weight¶

A_in¶

user_embedding¶

entity_embedding¶

relation_embedding¶

trans_w¶

aggregator_layers¶

tanh¶

mf_loss¶

reg_loss¶

restore_user_e = None¶

restore_entity_e = None¶

other_parameter_name = ['restore_user_e', 'restore_entity_e']¶

init_graph()[source]¶

Get the initial attention matrix through the collaborative knowledge graph

Returns:: Sparse tensor of the attention matrix
Return type:: torch.sparse.FloatTensor

_get_ego_embeddings()[source]¶

forward()[source]¶

_get_kg_embedding(h, r, pos_t, neg_t)[source]¶

calculate_loss(interaction)[source]¶

Calculate the training loss for a batch data.

Parameters:: interaction (Interaction) – Interaction class of the batch.
Returns:: Training loss, shape: []
Return type:: torch.Tensor

calculate_kg_loss(interaction)[source]¶

Calculate the training loss for a batch data of KG.

Parameters:: interaction (Interaction) – Interaction class of the batch.
Returns:: Training loss, shape: []
Return type:: torch.Tensor

generate_transE_score(hs, ts, r)[source]¶

Calculating scores for triples in KG.

Parameters:

hs (torch.Tensor) – head entities
ts (torch.Tensor) – tail entities
r (int) – the relation id between hs and ts

Returns:

the scores of (hs, r, ts)

Return type:

torch.Tensor

update_attentive_A()[source]¶: Update the attention matrix using the updated embedding matrix

predict(interaction)[source]¶

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

full_sort_predict(interaction)[source]¶

Full sort prediction function. Given users, calculate the scores between users and all candidate items.

Parameters:: interaction (Interaction) – Interaction class of the batch.
Returns:: Predicted scores for given users and all candidate items, shape: [n_batch_users * n_candidate_items]
Return type:: torch.Tensor