KDD-2013概述

精力有限，希望同学们补充“摘要”，以及划重点(加粗)

1. Document and topic models
   
   1. One Theme in All Views: Modeling Consensus Topics in Multiple Contexts
      
   2. Representing Documents Through Their Readers
      
      1. 以往文章研究的是给定文章model用户，这篇文章要给定用户model文章，比如一个用户的profile里有‘素食’，那么可能这个人分享的文章就和‘素食’有关。
         
      2. 构建了一个单词和类别的dictionary，然后利用单词线性组合成句子，来model文章
         
   3. Text-Based Measures of Document Diversity
      
   4. Diversity Maximization Under Matroid Constraints
      
      1. 文章聚类显示时，每个分类要多样性展示
         1. 定义距离函数，要求pair-wise的距离大
         2. 每个子分类只能有一个选入
            
2. Social media
   
   1. Connecting Users across Social Media Sites: A Behavioral-Modeling Approach
      
      1. cross-media user identication problem
         
   2. Automatic selection of social media responses to news
      
      1. 在所有refer某一篇article的‘tweet’中寻找最有代表性的，可以是视频之类的
         
   3.  Estimating Unbiased Sharer Reputation via Social Data Calibration
      
      1. not influential user, but 优质用户
         
   4. Linking Named Entities in Tweets with Knowledge Base via User Interest Modeling
      
      1. tweet entity linking
         
      2. graph model, each user has a underline topic distribution
         
3. Big data frameworks
   
   1. 暂不关注
4. Graph mining
   
   1. Selective Sampling on Graphs for Classification
      
      1. online learning
         
      2. graph sampling
5. Classification
   
   1. Density-Based Logistic Regression
      
      1. nonlinear logistic regression
         
      2. map the data to a feature space based on kernel density estimation
         
   2. MILS: Multi-Instance Learning from Multiple Information Sources
      
   3. Querying Discriminative and Representative Samples for Batch Mode Active Learning
      
   4. SVM_{pAUC}^{tight}: A New Support Vector Method for Optimizing Partial AUC Based on a Tight Convex Upper Bound
      
6. Healthcare and bioinformatics
   
   1. 暂不关心
7. Recommender systems
   
   1. Learning to question: Leveraging user preferences for shopping advice
      
      1. 生成一棵决策树，但是每个node是一个非专业性的问题（如买电脑，问是否是学生？），在每个node都会生成一个ranklist，可以在任意层退出
         
   2. Active Learning and Search on Low-Rank Matrices
      
   3. LCARS: A Location-Content-Aware Recommender System
      
8. Temporal/social influence
   
   1. Discovering Latent Influence in Online Social Activities via Shared Cascade Poisson Processes
      
      1. shared cascade Poisson processes for discovering latent influence from multiple sequences of item adoption events
         
   2. STRIP: Stream Learning of Influence Probabilities
      
   3. Fast Structure Learning in Generalized Stochastic Processes with Latent Factors
      
9. Sparse learning
   
   1. Robust Sparse Estimation of Multiresponse Regression and Inverse Covariance Matrix via the L2 distance
      
   2. Exact Sparse Recovery with L0 Projections
      
   3. Robust Principal Component Analysis via Capped Norms