recome_wan

• 1.本项目通过pytorch 框架大概复现了推荐系统相关的22个模型，其中包含多篇排序论文和多篇序列召回和图召回论文。
  
• 2.本项目一共包含demo和recome_wan这两个文件夹。
  
• 3.在demo里面分别选了一个召回模型和排序模型来作为示例，如果想调试其他的召回和排序模型，可以直接修改demo里面的rank_example.py文件或者recall_example.py的代码即可。
  
• 4.在recome_wan这个文件夹里，一共包含datasets、models、trainer、utils这四个大的模块。
  
• 5.其中datasets文件夹主要是数据类型和数据编码的处理，models里面包含了layers、rank_models和recall_models这三个文件夹。layers主要存放的是一些通用的层比如embedding层、Mlp层。 rank_models里存放的就是排序相关的模型，recall_models里存放的就是召回相关的模型。trainer主要是用来训练、验证、测试召回和排序的模型。utils包含一些关于召回模型的评价。 注：以下表格只展示了一些重要的模型，对一些简单的模型不做表格展示。

model	paper
dcn	DCN:Deep & Cross Network for Ad Click Predictions
deepfm	DeepFM: A Factorization-Machine based Neural Network for CTR Prediction
fibinet	FiBiNET: Combining Feature Importance and Bilinear feature Interaction for Click-Through Rate Prediction
mmoe	MMOE:Modeling Task Relationships in Multi-task Learning with Multi-gate Mixture-of-Experts
line	LINE: Large-scale Information Network Embedding
comirec	COMIREC:Controllable Multi-Interest Framework for Recommendation
gru4rec	GRU4REC:Session-based Recommendations with Recurrent Neural Networks
mind	MIND:Multi-interest network with dynamic routing for recommendation at Tmall
youtubednn	YOUTUBEDNN:Deep Neural Networks for YouTube Recommendations
lightgcn	LightGcn: Simplifying and Powering Graph Convolution Network for Recommendation
deepwalk	DeepWalk: Online Learning of Social Representations
node2vec	Node2Vev: Scalable Feature Learning for Networks
din	DIN:Deep Interest Network for Click-Through Rate Prediction
sdne	SDNE:Structural Deep Network Embedding
graphsage	GRAPHSAGE:Inductive Representation Learning on Large Graphs
eges	EGES:Enhanced Graph Embedding with Side Information