RepeatNet

Reference:

Pengjie Ren et al. “RepeatNet: A Repeat Aware Neural Recommendation Machine for Session-based Recommendation.” in AAAI 2019

Reference code:

https://github.com/PengjieRen/RepeatNet.

class recbole.model.sequential_recommender.repeatnet.Explore_Recommendation_Decoder(hidden_size, seq_len, num_item, device, dropout_prob)

Bases: torch.nn.modules.module.Module

forward(all_memory, last_memory, item_seq, mask=None)

calculate the force of explore

training: bool

class recbole.model.sequential_recommender.repeatnet.RepeatNet(config, dataset)

Bases: recbole.model.abstract_recommender.SequentialRecommender

RepeatNet explores a hybrid encoder with an repeat module and explore module repeat module is used for finding out the repeat consume in sequential recommendation explore module is used for exploring new items for recommendation

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(item_seq, item_seq_len)

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.

full_sort_predict(interaction)

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

input_type = 1

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

repeat_explore_loss(item_seq, pos_item)

training: bool

class recbole.model.sequential_recommender.repeatnet.Repeat_Explore_Mechanism(device, hidden_size, seq_len, dropout_prob)

Bases: torch.nn.modules.module.Module

forward(all_memory, last_memory)

calculate the probability of Repeat and explore

training: bool

class recbole.model.sequential_recommender.repeatnet.Repeat_Recommendation_Decoder(device, hidden_size, seq_len, num_item, dropout_prob)

Bases: torch.nn.modules.module.Module

forward(all_memory, last_memory, item_seq, mask=None)

calculate the the force of repeat

training: bool

recbole.model.sequential_recommender.repeatnet.build_map(b_map, device, max_index=None)

project the b_map to the place where it in should be like this:

item_seq A: [3,4,5] n_items: 6

after map: A

[0,0,1,0,0,0]

[0,0,0,1,0,0]

[0,0,0,0,1,0]

batch_size * seq_len ==>> batch_size * seq_len * n_item

use in RepeatNet:

[3,4,5] matmul [0,0,1,0,0,0]

[0,0,0,1,0,0]

[0,0,0,0,1,0]

==>>> [0,0,3,4,5,0] it works in the RepeatNet when project the seq item into all items

batch_size * 1 * seq_len matmul batch_size * seq_len * n_item ==>> batch_size * 1 * n_item