动态更新工作中实现或者阅读过的计算广告相关论文、学习资料和业界分享,作为自己工作的总结,也希望能为计算广告相关行业的同学带来便利。 所有资料均来自于互联网,如有侵权,请联系_王喆_。同时欢迎对计算广告感兴趣的同学与我讨论相关问题,我的联系方式如下:
- Email: [email protected]
- LinkedIn: 王喆的LinkedIn
- 知乎私信: 王喆的知乎
会不断加入一些重要的计算广告相关论文和资料,并去掉一些过时的或者跟计算广告不太相关的论文
New!
Image Matters- Visually modeling user behaviors using Advanced Model Server.pdf
阿里提出引入商品图像特征的(Deep Image CTR Model)CTR预估模型,并介绍其分布式机器学习框架AMSNew!
Deep Interest Network for Click-Through Rate Prediction.pdf
阿里提出的深度兴趣网络(Deep Interest Network)CTR预估模型
其他相关资源
- 张伟楠的RTB Papers列表
- 基于Spark MLlib的CTR预估模型(LR, FM, RF, GBDT, NN, PNN)
- 推荐系统相关论文和资源列表
- Honglei Zhang的推荐系统论文列表
Online Optimization,Parallel SGD,FTRL等优化方法,实用并且能够给出直观解释的文章
- Google Vizier A Service for Black-Box Optimization.pdf
Google的深度学习自动调参框架Vizier - 在线最优化求解(Online Optimization)-冯扬.pdf
非常推荐冯扬的这个教程,把在线优化问题讲的非常透 - Hogwild A Lock-Free Approach to Parallelizing Stochastic Gradient Descent.pdf
- Parallelized Stochastic Gradient Descent.pdf
- A Survey on Algorithms of the Regularized Convex Optimization Problem.pptx
- Follow-the-Regularized-Leader and Mirror Descent- Equivalence Theorems and L1 Regularization.pdf
- A Review of Bayesian Optimization.pdf
- Taking the Human Out of the Loop- A Review of Bayesian Optimization.pdf
- 非线性规划.doc
作为计算广告的核心,CTR预估永远是研究的热点,下面每一篇都是非常流行的文章,推荐逐一精读
- Deep Crossing- Web-Scale Modeling without Manually Crafted Combinatorial Features.pdf
- Learning Piece-wise Linear Models from Large Scale Data for Ad Click Prediction.pdf
阿里提出的Large Scale Piece-wise Linear Model (LS-PLM) CTR预估模型 - [GBDT+LR]Practical Lessons from Predicting Clicks on Ads at Facebook.pdf
- [FNN]Deep Learning over Multi-field Categorical Data.pdf
- Entire Space Multi-Task Model_ An Effective Approach for Estimating Post-Click Conversion Rate.pdf
- Deep Interest Network for Click-Through Rate Prediction.pdf
- Bid-aware Gradient Descent for Unbiased Learning with Censored Data in Display Advertising.pdf
RTB 中训练 CTR 模型数据集是赢得出价的广告,预测时的样本却是所有候选的广告,也就是训练集和测试集的分布不一致,这篇文章就是要消除这样的 bias - [Multi-Task]An Overview of Multi-Task Learning in Deep Neural Networks.pdf
- Ad Click Prediction a View from the Trenches.pdf
Google大名鼎鼎的用FTRL解决CTR在线预估的工程文章,非常经典。 - [PNN]Product-based Neural Networks for User Response Prediction.pdf
- Image Matters- Visually modeling user behaviors using Advanced Model Server.pdf
阿里提出引入商品图像特征的(Deep Image CTR Model)CTR预估模型,并介绍其分布式机器学习框架 Advanced Model Server (AMS) - [Wide & Deep]Wide & Deep Learning for Recommender Systems.pdf
- [DeepFM]- A Factorization-Machine based Neural Network for CTR Prediction.pdf
- Logistic Regression in Rare Events Data.pdf
样本稀少情况下的LR模型训练,讲的比较细 - Deep & Cross Network for Ad Click Predictions.pdf
Google 在17年发表的 Deep&Cross 网络,类似于 Wide&Deep, 比起 PNN 只做了特征二阶交叉,Deep&Cross 理论上能够做任意高阶的特征交叉 - Learning Deep Structured Semantic Models for Web Search using Clickthrough Data.pdf
- Adaptive Targeting for Online Advertisement.pdf
一篇比较简单但是全面的CTR预估的文章,有一定实用性
话题模型相关文章,PLSA,LDA,进行广告Context特征提取,创意优化经常会用到Topic Model
- 概率语言模型及其变形系列.pdf
- Parameter estimation for text analysis.pdf
- LDA数学八卦.pdf
- Distributed Representations of Words and Phrases and their Compositionality.pdf
- Dirichlet Distribution, Dirichlet Process and Dirichlet Process Mixture(PPT).pdf
- 理解共轭先验.pdf
Google三大篇,HDFS,MapReduce,BigTable,奠定大数据基础架构的三篇文章,任何从事大数据行业的工程师都应该了解
- MapReduce Simplified Data Processing on Large Clusters.pdf
- The Google File System.pdf
- Bigtable A Distributed Storage System for Structured Data.pdf
FM因子分解机模型的相关paper,在计算广告领域非常实用的模型
- FM PPT by CMU.pdf
- Field-aware Factorization Machines for CTR Prediction.pdf
- Factorization Machines Rendle2010.pdf
- libfm-1.42.manual.pdf
- Scaling Factorization Machines to Relational Data.pdf
- Fast Context-aware Recommendations with Factorization Machines.pdf
- fastFM- A Library for Factorization Machines.pdf
广告系统中Pacing,预算控制,以及怎么把预算控制与其他模块相结合的问题
- Budget Pacing for Targeted Online Advertisements at LinkedIn.pdf
linkedin的一篇非常有工程价值的解决pacing问题的文章,强烈建议计算广告系统采用此方法。 - Predicting Traffic of Online Advertising in Real-time Bidding Systems from Perspective of Demand-Side Platforms.pdf
- Real Time Bid Optimization with Smooth Budget Delivery in Online Advertising.pdf
如何将Pcaing与效果优化结合在一起,这篇文章讲的很清楚 - PID控制经典培训教程.pdf
PID控制的经典教程 - PID控制原理与控制算法.doc
对于采用PID控制解决pacing问题,该文章是PID控制原理比较清晰的介绍文章。 - Smart Pacing for Effective Online Ad Campaign Optimization.pdf
跟上篇文章一样,都是雅虎同一组人写的,解决预算控制与效果结合的问题,可以跟上篇文章一起看了
树模型和基于树模型的boosting模型,树模型的效果在大部分问题上非常好,在CTR,CVR预估及特征工程方面的应用非常广
- Introduction to Boosted Trees.pdf
- Classification and Regression Trees.pdf
- Greedy Function Approximation A Gradient Boosting Machine.pdf
- Classification and Regression Trees.ppt
事实上,现在很多大的媒体主仍是合约广告系统,合约广告系统的在线分配,Yield Optimization,以及定价问题都是非常重要且有挑战性的问题
- A Dynamic Pricing Model for Unifying Programmatic Guarantee and Real-Time Bidding in Display Advertising.pdf
- Pricing Guaranteed Contracts in Online Display Advertising.pdf
- Risk-Aware Dynamic Reserve Prices of Programmatic Guarantee in Display Advertising.pdf
- Pricing Guidance in Ad Sale Negotiations The PrintAds Example.pdf
- Risk-Aware Revenue Maximization in Display Advertising.pdf
计算广告中广告定价,RTB过程中广告出价策略的相关问题
- Research Frontier of Real-Time Bidding based Display Advertising.pdf
张伟楠博士的一篇介绍竞价算法的ppt,可以非常清晰的了解该问题的主要方法 - Budget Constrained Bidding by Model-free Reinforcement Learning in Display Advertising.pdf
- Real-Time Bidding with Multi-Agent Reinforcement Learning in Display Advertising.pdf
- Real-Time Bidding by Reinforcement Learning in Display Advertising.pdf
- Combining Powers of Two Predictors in Optimizing Real-Time Bidding Strategy under Constrained Budget.pdf
国立台湾大学的文章,介绍一种基于流量选择的计算广告竞价方法,有别于传统的CTR CPC的方法,我在实践中尝试过该方法,非常有效 - Real-Time Bidding Algorithms for Performance-Based Display Ad Allocation.pdf
微软的一篇基于PID反馈控制的与效果相关的竞价算法 - Deep Reinforcement Learning for Sponsored Search Real-time Bidding.pdf
阿里妈妈搜索广告团队的论文,通过 Reinforcement Learning 探索实时出价问题 - Optimized Cost per Click in Taobao Display Advertising.pdf
淘宝搜索广告中 OCPC 模式的出价策略
广告系统的架构问题
- Parameter Server for Distributed Machine Learning.pdf
- [TensorFlow Whitepaper]TensorFlow- Large-Scale Machine Learning on Heterogeneous Distributed Systems.pdf
- 大数据下的广告排序技术及实践.pdf
阿里妈妈的一篇广告排序问题的ppt,模型、训练、评估都有涉及,很有工程价值 - 美团机器学习 吃喝玩乐中的算法问题.pdf
美团王栋博士的一篇关于美团机器学习相关问题的介绍,介绍的比较全但比较粗浅,可以借此了解美团的一些机器学习问题 - Display Advertising with Real-Time Bidding (RTB) and Behavioural Targeting.pdf
张伟楠博士的RTB过程所有相关算法的书,全而精,非常棒 - A Comparison of Distributed Machine Learning Platforms.pdf
- Efficient Query Evaluation using a Two-Level Retrieval Process.pdf
搜索广告中经典的搜索算法 Wand(Weak AND) - [TensorFlow Whitepaper]TensorFlow- A System for Large-Scale Machine Learning.pdf
- Scaling Distributed Machine Learning with the Parameter Server.pdf
- Overlapping Experiment Infrastructure More, Better, Faster Experimentation.pdf
Google 一篇关于 A/B 测试框架的论文,涉及到如何切分流量以同时进行多个 A/B 测试,工程性很强
机器学习方面一些非常实用的学习资料
- 各种回归的概念学习.doc
- 机器学习总图.jpg
- Efficient Estimation of Word Representations in Vector Space.pdf
- Rules of Machine Learning- Best Practices for ML Engineering.pdf
- An introduction to ROC analysis.pdf
- Deep Learning Tutorial.pdf
- 广义线性模型.ppt
- 贝叶斯统计学(PPT).pdf
- 关联规则基本算法及其应用.doc
迁移学习相关文章,计算广告中经常遇到新广告冷启动的问题,利用迁移学习能较好解决该问题
探索和利用,计算广告中非常经典,也是容易被大家忽视的问题,其实所有的广告系统都面临如何解决新广告主冷启动,以及在效果不好的情况下如何探索新的优质流量的问题,希望该目录下的几篇文章能够帮助到你
- An Empirical Evaluation of Thompson Sampling.pdf
- Dynamic Online Pricing with Incomplete Information Using Multi-Armed Bandit Experiments.pdf
- Finite-time Analysis of the Multiarmed Bandit Problem.pdf
- A Fast and Simple Algorithm for Contextual Bandits.pdf
- Customer Acquisition via Display Advertising Using MultiArmed Bandit Experiments.pdf
- Mastering the game of Go with deep neural networks and tree search.pdf
- Exploring compact reinforcement-learning representations with linear regression.pdf
- Incentivizting Exploration in Reinforcement Learning with Deep Predictive Models.pdf
- Bandit Algorithms Continued- UCB1.pdf
- A Contextual-Bandit Approach to Personalized News Article Recommendation(LinUCB).pdf
- Exploitation and Exploration in a Performance based Contextual Advertising System.pdf
- Bandit based Monte-Carlo Planning.pdf
- Random Forest for the Contextual Bandit Problem.pdf
- Unifying Count-Based Exploration and Intrinsic Motivation.pdf
- Analysis of Thompson Sampling for the Multi-armed Bandit Problem.pdf
- Thompson Sampling PPT.pdf
- Hierarchical Deep Reinforcement Learning- Integrating Temporal Abstraction and Intrinsic Motivation.pdf
- Exploration and Exploitation Problem by Wang Zhe.pptx
- Exploration exploitation in Go UCT for Monte-Carlo Go.pdf
- 对抗搜索、多臂老虎机问题、UCB算法.ppt
- Using Confidence Bounds for Exploitation-Exploration Trade-offs.pdf
广告流量的分配问题
- An Efficient Algorithm for Allocation of Guaranteed Display Advertising.pdf
同样是雅虎的流量分配文章,跟上一篇文章同时发布,介绍SHALE流量分配算法 - Ad Serving Using a Compact Allocation Plan.pdf
雅虎的一篇比较经典的流量分配的文章,文中的HWM和DUAL算法都比较实用