1 Jie Tang 2 Jimeng Sun 3 Quan Lin 4 Fengjiao Wang 1 Department of Computer Science and Technology Tsinghua University China 2 IBM TJ Watson Research Center USA 3 ID: 437355
Download Presentation The PPT/PDF document "1 Chenhao Tan," is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
1Chenhao Tan, 1Jie Tang, 2Jimeng Sun, 3Quan Lin, 4Fengjiao Wang1Department of Computer Science and Technology, Tsinghua University, China2IBM TJ Watson Research Center, USA3Huazhong University of Science and Technology, China4Beijing University of Aeronautics and Astronautics, China
Social Action Tracking via Noise Tolerant Time-varying Factor GraphsSlide2
OutlineMotivationApproachExperimentConclusion & Future WorkSlide3
Motivation 500 million users the 3rd largest “Country” in the world More visitors than Google Action: Update statues, create event More than 4 billion imagesAction: Add tags, Add favorites 2009, 2 billion tweets per quarter 2010, 4 billion tweets per quarterAction: Post tweets, RetweetSlide4
User Action in Social NetworksAdd photo to her favoritesPost tweets on “Haiti Earthquake”Publish in KDD ConferenceTwitterFlickrArnetminerSlide5
User Action in Social NetworksQuestions:What factors influence you to add a photo into your favorite list?- If you post a tweet on “Haiti Earthquake”, will your friends retweet it or reply?Challenge: - How to track and model users’ actions? - How to predict users’ actions over time?Slide6
JohnTime tJohnTime t+1Action Prediction
:Will John post a tweet on “Haiti Earthquake”?Attributes:Always watch news
Enjoy sports
….
Influence
1
Personal attributes
4
Dependence
2
Complex Factors
Correlation
3Slide7
Problem formulationGt =(Vt, Et, Xt, Yt)
Nodes at time
t
Edges at time
t
Attribute matrix at time t
Actions at time
tSlide8
NTT-FGM ModelContinuous latent action statePersonal attributes
Correlation
Dependence
Influence
Action
Personal attributes Slide9
How to estimate the parameters?Model InstantiationSlide10
Model LearningExtremely time costing!!Our solution: distributed learning (MPI)Slide11
Most Time-costingCompute the gradientsDistributed LearningSlide12
Data SetBaselineSVMwvRN (Macskassy, 2003)Evaluation Measure:Precision, Recall, F1-MeasureActionNodes
#EdgesAction Stats
Twitter
Post tweets on
“
Haiti Earthquake”
7,521
304,275
730,568
Flickr
Add photos into favorite
list
8,721
485,253
485,253
Arnetminer
Issue
publications on KDD
2,062
34,986
2,960
ExperimentSlide13
Performance AnalysisSlide14
Factor Contribution Analysis NTT-FGM: Our model NTT-FGM-I: Our model ignoring influence NTT-FGM-CI: Our model ignoring influence and correlationSlide15
Efficiency PerformanceSlide16
ConclusionFormally formulate the problem of social action trackingPropose a unified model: NTT-FGM to simultaneously model various factorsPresent an efficient learning algorithm and develop a distributed implementation Validate the proposed approach on three different data sets, and our model achieves a better performanceSlide17
Thank you!QA?Data & Code: http://arnetminer.org/stntWelcome to our poster! Slide18
Statistical Study: InfluenceY-axis: the likelihood that the user also performs the action at tX-axis: the percentage of one’s friends who perform an action at t − 1Slide19
Statistical Study: DependenceY-axis: the likelihood that a user performs an actionX-axis: different time windowsSlide20
Statistical Study: CorrelationY-axis: the likelihood that two friends(random) perform an action togetherX-axis: different time windowsSlide21
AppendixSlide22
AppendixSlide23
AppendixSlide24
PredictionBased on the learning parameters we just need to solve the following equations:Slide25
Latent State AnalysisAction Bias Factor: f(y12|z12)Influence Factor: g(z11,z12)Correlation Factor: h(z12,
z22), h(z12,
x
1
2
)