# MCCF Source code for AAAI2020 paper ["**Multi-Component Graph Convolutional Collaborative Filtering**"](https://arxiv.xilesou.top/abs/1911.10699) ## Environment Settings * Python == 3.6.9 * torchvision == 0.4.2 * numpy == 1.17.3 * scikit-learn == 0.21.3 ## Parameter Settings - epochs: the number of epochs to train 训练的迭代次数 - lr: learning rate 学习率 - embed_dim: embedding dimension 嵌入尺寸 - N: a parameter of L0, the default is the number of triples 参数为L0，默认为三组数 - droprate: dropout rate 下坠率 - batch_size: batch size for training 训练的批处理大小 ## Files in the folder ~~~~ MCCF/ ├── run.py: training the model 训练模型 ├── utils/ │ ├── aggregator.py: aggregating the feature of neighbors 聚合邻居的特征 │ ├── l0dense.py: implementation of L0 regularization for a fully connected layer 完整连接层的L0正则化的实现 │ ├── attention.py: implementation of the node-level attention 节点级注意力的实现 │ ├── encoder.py: together with aggregator to form the decomposer 与聚合器一起组成分解器 │ └── combiner.py: implementation of the combiner 合成器的实现 ├── datasets/ │ ├── yelp/ │ │ ├── business_user.txt │ │ ├── preprocess.py: data preprocessing example 数据预处理示范 │ │ └── _allData.p │ ├── amazon/ │ │ ├── user_item.dat │ │ └── _allData.p │ └── movielens/ │ ├── ub.base │ ├── ub.test │ ├── ua.base │ ├── ua.test │ ├── u5.base │ ├── u5.test │ ├── u4.base │ ├── u4.test │ ├── u3.base │ ├── u3.test │ ├── u2.base │ ├── u2.test │ ├── u1.base │ ├── u1.test │ ├── u.data │ ├── u.user │ ├── u.item │ └── _allData.p └── README.md ~~~~ ## Data ### Input training data 输入训练数据 * u_adj: user's purchased history (item set in training set) 用户购买历史(训练集中的项目集) * i_adj: user set (in training set) who have interacted with the item 与项目交互的用户集(在训练集中) * u_train, i_train, r_train: training set (user, item, rating) 训练集(用户，项目，等级) * u_test, i_test, r_test: testing set (user, item, rating) 测试集(用户，项目，等级) ### Input pre-trained data 输入预先训练过的数据 * u2e, i2e: for small data sets, the corresponding vectors in the rating matrix can be used as initial embeddings; for large data sets, we recommend using the embeddings of other models (e.g., GC-MC) as pre-training, which greatly reduces the complexity. 对于小数据集，可以使用评级矩阵中对应的向量作为初始嵌入;对于大数据集，我们建议使用其他模型的嵌入(如GC-MC)作为预训练，这样可以大大降低复杂性。 ## Basic Usage ~~~ python run.py ~~~ HINT: the sampling thresholds in aggregator.py change with dataset. 注：采样阈值随数据集的变化而变化。 # Reference ``` @article{wang2019multi, title={Multi-Component Graph Convolutional Collaborative Filtering}, author={Wang, Xiao and Wang, Ruijia and Shi, Chuan and Song, Guojie and Li, Qingyong}, journal={arXiv preprint arXiv:1911.10699}, year={2019} } ```