Source code for recbole.model.sequential_recommender.caser

# -*- coding: utf-8 -*-
# @Time   : 2020/9/21
# @Author : Jingsen Zhang
# @Email  : zhangjingsen@ruc.edu.cn

# UPDATE
# @Time   : 2020/10/2
# @Author : Jingsen Zhang
# @Email  : zhangjingsen@ruc.edu.cn

r"""
Caser
################################################

Reference:
    Jiaxi Tang et al., "Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding" in WSDM 2018.

Reference code:
    https://github.com/graytowne/caser_pytorch

"""

import torch
from torch import nn
from torch.nn import functional as F
from torch.nn.init import normal_, xavier_normal_, constant_

from recbole.model.abstract_recommender import SequentialRecommender
from recbole.model.loss import RegLoss, BPRLoss


[docs]class Caser(SequentialRecommender): r"""Caser is a model that incorporate CNN for recommendation. Note: We did not use the sliding window to generate training instances as in the paper, in order that the generation method we used is common to other sequential models. For comparison with other models, we set the parameter T in the paper as 1. In addition, to prevent excessive CNN layers (ValueError: Training loss is nan), please make sure the parameters MAX_ITEM_LIST_LENGTH small, such as 10. """ def __init__(self, config, dataset): super(Caser, self).__init__(config, dataset) # load parameters info self.embedding_size = config['embedding_size'] self.loss_type = config['loss_type'] self.n_h = config['nh'] self.n_v = config['nv'] self.dropout_prob = config['dropout_prob'] self.reg_weight = config['reg_weight'] # load dataset info self.n_users = dataset.user_num # define layers and loss self.user_embedding = nn.Embedding(self.n_users, self.embedding_size, padding_idx=0) self.item_embedding = nn.Embedding(self.n_items, self.embedding_size, padding_idx=0) # vertical conv layer self.conv_v = nn.Conv2d(in_channels=1, out_channels=self.n_v, kernel_size=(self.max_seq_length, 1)) # horizontal conv layer lengths = [i + 1 for i in range(self.max_seq_length)] self.conv_h = nn.ModuleList([ nn.Conv2d(in_channels=1, out_channels=self.n_h, kernel_size=(i, self.embedding_size)) for i in lengths ]) # fully-connected layer self.fc1_dim_v = self.n_v * self.embedding_size self.fc1_dim_h = self.n_h * len(lengths) fc1_dim_in = self.fc1_dim_v + self.fc1_dim_h self.fc1 = nn.Linear(fc1_dim_in, self.embedding_size) self.fc2 = nn.Linear(self.embedding_size + self.embedding_size, self.embedding_size) self.dropout = nn.Dropout(self.dropout_prob) self.ac_conv = nn.ReLU() self.ac_fc = nn.ReLU() self.reg_loss = RegLoss() if self.loss_type == 'BPR': self.loss_fct = BPRLoss() elif self.loss_type == 'CE': self.loss_fct = nn.CrossEntropyLoss() else: raise NotImplementedError("Make sure 'loss_type' in ['BPR', 'CE']!") # parameters initialization self.apply(self._init_weights) def _init_weights(self, module): if isinstance(module, nn.Embedding): normal_(module.weight.data, 0, 1.0 / module.embedding_dim) elif isinstance(module, nn.Linear): xavier_normal_(module.weight.data) if module.bias is not None: constant_(module.bias.data, 0)
[docs] def forward(self, user, item_seq): # Embedding Look-up # use unsqueeze() to get a 4-D input for convolution layers. (batch_size * 1 * max_length * embedding_size) item_seq_emb = self.item_embedding(item_seq).unsqueeze(1) user_emb = self.user_embedding(user).squeeze(1) # Convolutional Layers out, out_h, out_v = None, None, None # vertical conv layer if self.n_v: out_v = self.conv_v(item_seq_emb) out_v = out_v.view(-1, self.fc1_dim_v) # prepare for fully connect # horizontal conv layer out_hs = list() if self.n_h: for conv in self.conv_h: conv_out = self.ac_conv(conv(item_seq_emb).squeeze(3)) pool_out = F.max_pool1d(conv_out, conv_out.size(2)).squeeze(2) out_hs.append(pool_out) out_h = torch.cat(out_hs, 1) # prepare for fully connect # Fully-connected Layers out = torch.cat([out_v, out_h], 1) # apply dropout out = self.dropout(out) # fully-connected layer z = self.ac_fc(self.fc1(out)) x = torch.cat([z, user_emb], 1) seq_output = self.ac_fc(self.fc2(x)) # the hidden_state of the predicted item, size:(batch_size * hidden_size) return seq_output
[docs] def reg_loss_conv_h(self): r""" L2 loss on conv_h """ loss_conv_h = 0 for name, parm in self.conv_h.named_parameters(): if name.endswith('weight'): loss_conv_h = loss_conv_h + loss_conv_h * parm.norm(2) return self.reg_weight * loss_conv_h
[docs] def calculate_loss(self, interaction): item_seq = interaction[self.ITEM_SEQ] user = interaction[self.USER_ID] seq_output = self.forward(user, item_seq) pos_items = interaction[self.POS_ITEM_ID] if self.loss_type == 'BPR': neg_items = interaction[self.NEG_ITEM_ID] pos_items_emb = self.item_embedding(pos_items) neg_items_emb = self.item_embedding(neg_items) pos_score = torch.sum(seq_output * pos_items_emb, dim=-1) # [B] neg_score = torch.sum(seq_output * neg_items_emb, dim=-1) # [B] loss = self.loss_fct(pos_score, neg_score) else: # self.loss_type = 'CE' test_item_emb = self.item_embedding.weight logits = torch.matmul(seq_output, test_item_emb.transpose(0, 1)) loss = self.loss_fct(logits, pos_items) reg_loss = self.reg_loss([ self.user_embedding.weight, self.item_embedding.weight, self.conv_v.weight, self.fc1.weight, self.fc2.weight ]) loss = loss + self.reg_weight * reg_loss + self.reg_loss_conv_h() return loss
[docs] def predict(self, interaction): item_seq = interaction[self.ITEM_SEQ] user = interaction[self.USER_ID] test_item = interaction[self.ITEM_ID] seq_output = self.forward(user, item_seq) test_item_emb = self.item_embedding(test_item) scores = torch.mul(seq_output, test_item_emb).sum(dim=1) # [B] return scores
[docs] def full_sort_predict(self, interaction): item_seq = interaction[self.ITEM_SEQ] user = interaction[self.USER_ID] seq_output = self.forward(user, item_seq) test_items_emb = self.item_embedding.weight scores = torch.matmul(seq_output, test_items_emb.transpose(0, 1)) # [B, n_items] return scores