计算广告中常用模型的相关资料整理

计算广告中的常用模型

---

前言

  我会在这里动态更新我近期学习和工作中阅读过的计算广告相关论文和代码，作为自己学习的整理和总结。由于目前有逐个实现的计划，所以在这里构建一个传送门，方便之后实现的时候进行查阅。本文所有资料和链接均来自于互联网，直接搜索都能找得到。

---

传送门

MF（Matrix Factorization）
【github】nimfa, MF
【github】MF
【Paper】Matrix Factorization techniques for Recommender Systems
【Paper】Predicting movie ratings and recommender systems

MLR（多元线性回归模型）= Embedding + MF + LR
【Paper】Learning Piece-wise Linear Models from Large Scale Data for Ad Click Prediction (阿里提出)
【CSDN】MLR, Paper笔记

FM = LR + MF
【github】libfm
【github】pyFM, coreylynch
【github】FM, Johnson0722
【Paper】Factorization Machines

FFM
【github】libffm
【github】xlearn
【github】FFM, Johnson0722
【Paper】Field-aware Factorization Machines for CTR Prediction
【Blog】深入FFM原理与实践_美团
【PPT】3 Idiots’ Approach for Display Advertising Challenge

FNN = FM + MLP
【zhihu】FNN, nzc
【github】FNN, nzc, torch
【Paper】Deep Learning over Multi-Field Categorical Data: A Case Study on User Response Prediction

PNN = FNN + product layer
【zhihu】PNN, nzc
【github】PNN, nzc, torch
【github】PNN, zhougr1993, tensorflow
【github】PNN, Atomu2014, tensorflow
【github】PNN, lambdaji, tensorflow
【Paper】Product-based Neural Networks for User Response Prediction

WDL（Wide & Deep）= LR + Embedding + MLP
【github】 wide&deep, ichuang, tensorflow
【github】wide&deep, jrzaurin, keras
【github】wide&deep, jorahn, keras
【github】wide&deep, zhougr1993, tensorflow
【github】wide&deep, lambdaji, tensorflow
【API】wide&deep, tensorflow, tf-API
【Paper】Wide & Deep Learning for Recommender Systems

DeepFM = FM + Embedding + MLP
【github】deepfm, chenglongchen, tensorflow
【github】deepfm, Johnson0722, tensorflow
【github】deepfm, zgw21cn, tensorflow
【github】deepfm, nzc, torch
【github】deepfm, zhougr1993, tensorflow
【github】deepfm, lambdaji, tensorflow
【github】deepfm, shenweichen, tensorflow
【Paper】DeepFM: A Factorization-Machine based Neural Network for CTR Prediction

NFM（Neural FM） = FM + Embedding + MLP + Bi-Interaction Pooling
【zhihu】NFM, nzc
【github】NFM, nzc, torch
【github】NFM, lambdaji, tensorflow
【Paper】Neural Factorization Machines for Sparse Predictive Analytics

AFM（Attentional FM）= LR + Embedding + MLP + Attention
【zhihu】AFM, nzc
【github】AFM, lambdaji, tensorflow
【github】AFM, nzc, torch
【github】AFM, hexiangnan
【github】AFM, cnfsll, tensorflow
【Paper】Attentional Factorization Machines: Learning the Weight of Feature Interactions via Attention Networks
【Blog】Attention Mechanism

DIN（Deep Interest Network）= Embedding + MLP + Attention
【github】DIN, nzc, torch
【github】DIN, zhougr1993, tensorflow
【Paper】Deep Interest Network for Click-Through Rate Prediction (阿里提出)

DIEN (Deep Interest Evolution Network)
【github】DIEN, mouna99, tensorflow
【Paper】Deep Interest Evolution Network for Click-Through Rate Prediction

DeepCTR = CNN + Embedding + MLP
【Paper】Deep CTR Prediction in Display Advertising

DCN （Deep & Cross Network） = Embedding + ResNet + LR
【zhihu】DCN, nzc
【github】DCN, nzc, torch
【github】DCN, lambdaji, tensorflow
【github】DCN, shenweichen, tensorflow
【Paper】Deep & Cross Network for Ad Click Predictions

Deep Embedding Forest = Embedding + Forest
【Paper】Deep Embedding Forest: Forest-based Serving with Deep Embedding Features
【Paper】Deep Forest: Towards An Alternative to Deep Neural Networks
【github】deep forest, kingfengji

MVM
【github】DeepMVM, lambdaji, tensorflow
【zhihu】DeepMVM, lambdaji

XNN
【github】XNN, chenglongchen

---

其他

CTR预估算法之FM, FFM, DeepFM及实践
用户在线广告点击行为预测的深度学习模型（含PPT下载）
闲聊DNN CTR预估模型
常见计算广告点击率预估算法总结
计算广告Papers汇总
从ctr预估问题看看f(x)设计—DNN篇
TensorFlow Estimator of Deep CTR --DeepFM/NFM/AFM/FNN/PNN
解析微信朋友圈的lookalike算法
腾讯社交广告 Lookalike 相似人群拓展定向使用指南