1 / 38

Goals for Today’s Meeting

Scalable Methods for the Analysis of Network-Based Data MURI Project: University of California, Irvine Principal Investigator: Padhraic Smyth Kick-off Meeting November 18 th 2008. Goals for Today’s Meeting. Review overall goals and research of MURI project University research groups

cianna
Download Presentation

Goals for Today’s Meeting

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Scalable Methods for the Analysis of Network-Based DataMURI Project: University of California, Irvine Principal Investigator: Padhraic Smyth Kick-off MeetingNovember 18th 2008

  2. Goals for Today’s Meeting • Review overall goals and research of MURI project • University research groups • learn about each other’s research • See opportunities for collaboration • MURI team and ONR/Navy • MURI team: learn about ONR interests • ONR: learn about expertise and plans of MURI team • Action items • Future meetings and collaborative activities • Review Year 1 research goals Butts

  3. Outline • Introductions • Review today’s agenda • Schedule of talks • Logistics • Overview of our MURI research project • Themes and goals • Tasks

  4. MURI Investigators Padhraic Smyth UCI David Eppstein UCI Carter Butts UCI Michael Goodrich UCI Mark Handcock U Washington Dave Mount U Maryland Dave Hunter Penn State

  5. MURI Project Participants • Postdocs • UC Irvine • Romain Thibaux (Computer Science) • Graduate Students (all UC Irvine) • Computer Science • Darren Strash • Lowell Trott • Statistics • Chris DuBois • Social Science • Chris Marcum • Lorien Jasny • Emma Spiro • Zack Almquist

  6. Outline • Introductions • Review today’s agenda • Schedule of talks • Logistics • Overview of our MURI research project • Themes and goals • Tasks

  7. Agenda for MURI Kickoff Meeting at UC Irvine November 18th 2008Location: UC Irvine, Bren Hall, Room 4011 MORNING SESSION8:30  Arrive, continental breakfast available9:00  Introductions and overview of MURI proposal         Padhraic Smyth (UCI)9:30  Research and application challenges from ONR's perspective          Martin Kruger (ONR, MURI Program Manager)10:00 Brief discussion/Q&A session between ONR representatives and PIs

  8. Agenda for MURI Kickoff Meeting at UC Irvine 10:15 Break10:30  Tutorial Session: Statistical models for network data:       Mark Handcock (U Washington), Carter Butts (UCI), Dave Hunter (Penn State)        - Fundamentals of exponential family random graph models (ERGMs)          - Parameter estimation in ERGMs: principles and computational challenges          - Alternative statistical frameworks such as latent space models LUNCH12:00 Lunch for PIs and UCI visitors at the University Club (next door to Bren Hall)

  9. Agenda for MURI Kickoff Meeting at UC Irvine AFTERNOON SESSION: 15-minute brief Research Presentations: 1:30   - Studying networks through an algorithmic lens           Michael Goodrich (UCI) 1:45   - Fast algorithms for computing network statistics:           David Eppstein (UCI)2:00   - Data structures for dynamic and kinetic multidimensional point sets:           Dave Mount (U Maryland)2:15  - Modeling dynamic and relational event data:           Carter Butts2:30   - Statistical modeling of large text collectons:            Padhraic Smyth (UCI)2:45   - Modeling partially observed network data:           Mark Handcock (U Washington)

  10. Agenda for MURI Kickoff Meeting at UC Irvine 3:00  Break and Informal Discussion3:30  Brief talks on Software and Data Sets:         - R software for network analysis:           Dave Hunter (Penn State)          - Experimental results on real-world networks         PhD students from Carter Butt's group (Sociology Dept, UCI)         - Large network data sets for experimentation:          Chris DuBois (PhD student, Statistics Department, UCI)

  11. Agenda for MURI Kickoff Meeting at UC Irvine 4:15  Open Discussion on Research Plans        - relation of MURI research topics to military applications        - further opportunities for collaboration within the team        - year 1 research goals 5:00  Organizational Issues and Wrap-up          - future meetings (frequency, location)        - encouraging interaction between team members          (conference calls, weekly research meetings, etc)         - use of collaborative Web pages         - action items5:30  Adjourn     

  12. Logistics • Meals • Lunch at University Club - for PIs and non-UCI folks • Coffee breaks • Wireless • Should be able to get 24-hour guest access • Slides • Will be available online by the end of today

  13. Outline • Introductions • Review today’s agenda • Schedule of talks • Logistics • Overview of our MURI research project • Themes and goals • Tasks

  14. MURI Project: Scalable Methods for Analysis of Network-Based Data • 4 universities collaborating, 7 PIs • Support for (approx) 8 graduate students and 3 postdocs or research associates • 3-year project with possible extension to 5 years • Time Period • Funding arrived at UCI in September 2008 • At other universities in Sept/Oct 2008 • Official project start/end • June 1 2008 to May 30 2011/2013

  15. A Small Social Network Eppstein Butts Butts Hunter Mount Smyth Handcock Goodrich

  16. A Small Social Network Eppstein Butts Hunter Mount Smyth Handcock Goodrich Statistics

  17. A Small Social Network Eppstein Butts Hunter Mount Smyth Handcock Goodrich Statistics Algorithms & Data Structures

  18. Figure from Carter Butts

  19. Statistical Modeling of Network Data • Statistics = principled approach for inference from data • Basis for optimal prediction • querying = computation of conditional probabilities/expectation • Principles for handling noisy measurements • e.g., noisy edge observation process • Integration of different sources of information • e.g., combining edge information with node covariates • Quantification of uncertainty • e.g., which model is a better explanation of the data

  20. Limitations of Existing Methods • Network data over time • Relatively few statistical models for dynamic network data • Heterogeneous data • e.g., few techniques for incorporating text, spatial information, etc, into network models • Computational tractability • Many network modeling algorithms scale exponentially in the number of nodes N

  21. Example • G = {V, E} V = set of N nodes E = set of directed binary edges • Exponential random graph model (ERGM) P(G | q) = f( G ; q ) / normalization constant The normalization constant = sum over all possible graphs How many graphs? 2 N(N-1) e.g., N = 20, we have 2380 ~ 1038 graphs to sum over

  22. Key Themes of our MURI Project • Foundational research on new statistical estimation techniques for network data • e.g., principles of modeling with missing data • New algorithms for heterogeneous network data • Incorporating time, space, text, other covariates • Faster algorithms • E.g., approximation methods for very large data sets • Software • Make network inference software publicly-available (in R)

  23. Key Themes of our MURI Project Fast Algorithms Statistical Methods Richer models Large Heterogeneous Data Sets Software

  24. Tasks A: Fast network estimation algorithms Eppstein, Butts B: Spatial representations and network data Goodrich, Eppstein, Mount C: Advanced network estimation techniques Handcock, Hunter D: Scalable methods for relational events Butts E: Network models with text data Smyth F: Software for network inference and prediction Hunter

  25. Task A: Fast Network Estimation Algorithms Investigators: Eppstein, Butts • Problem: • Statistical inference algorithms can be slow because of repeated computation of various statistics on graphs • Goal • Leverage ideas from computational graph algorithms to enable much faster computation – also enabling computation of more complex and realistic statistics • Projects • Dynamic graph methods for change-score computation • Rapid subgraph automorphism detection for feature counting • Dynamic connectivity

  26. Task B: Spatial Representations and Network Data Investigators: Goodrich, Eppstein, Mount • Problem: • Spatial representations of network data can be quite useful (both latent embeddings and actual spatial information) but current statistical modeling algorithms scale poorly • Goal • Build on recent efficient geometric data indexing techniques in computer science to develop much faster and efficient algorithms • Projects • Improved algorithms for latent-space embeddings • Fast implementations for high-dimensional latent space models • Techniques for integrating actual and latent space geometry

  27. Task C: Advanced Estimation Techniques Investigators: Handcock, Hunter • Problem: • Current statistical network inference models often make unrealistic assumptions, e.g., • Assume complete (non-missing) data • Assume that exact computation is possible • Goal • Develop new theories and techniques that relax these assumptions, i.e., methods for handing missing data and techniques for approximate inference • Projects • Inference with partially observed network data • Approximation methods • Approximate likelihood techniques • Approximate MCMC algorithms • Will leverage new techniques developed in Tasks A and B

  28. Figure from Dave Hunter, Penn State

  29. Task D: Scalable Temporal Models Investigator: Butts • Problem: • Few statistical methods for modeling temporal sequences of events among a network of actors • Goal • Develop new statistical relational event models to handle an evolving set of events over time in a network context • Projects • Specification of relational event statistics • Rapid likelihood computation for relational event models • Predictive event system queries • Interventions, forecasting, and “network steering” • Can build on ideas from Tasks A, B, C

  30. Task E: Network Models and Text Data Investigator: Smyth • Problem: • Lack of statistical techniques that can combine network and text data within a single framework (e.g., email communication) • Goal • Leverage recent advances in both statistical text mining and statistical network modeling to create new combined models • Projects • Latent variable models for text and network data • Text as exogenous data for statistical network models • Modeling of text and network data over time • Fast algorithms for statistical modeling of text/networks • Can build on ideas from Tasks A, B, C and D

  31. Network of email communication patterns in a corporate research lab

  32. Task F: Software for Network Inference and Prediction Investigator: Hunter • Goal • Disseminate algorithms and software to research and practitioner communities • How? • By incorporating our new algorithms into the R statistical package • R = open source language for stat computing/graphics • MURI team has significant prior experience with developing statistical network modeling packages in R • network (Butts et al, 2007) • latentnet (Handcock et al, 2004) • ergm(Handcock et al, 2003) • sna (Butts, 2000) • Will integrate algorithms and techniques from earlier tasks

  33. Data Sets • Traditional social network data sets • “Next generation” data sets • Dynamic network data • E.g., WTC communications • Network data with text • E.g., political blogs, Enron emails • Often much larger and richer than traditional data sets • See afternoon talks by PhD students Lorien Jasny and Chris DuBois

  34. Evaluation Methods • Traditional statistical metrics • Log-likelihood on training data • Model comparisons using penalized and marginal likelihood • Predictive metrics • E.g., for dynamic networks, prediction of edge and node properties “out of sample”, and assessment of the accuracy of these predictions • Classification accuracy, precision-recall (ROC), etc • Computational metrics • Worst and average-case analysis • Empirical evaluations of computation time • Trade-offs of statistical/predictive accuracy with computation time

  35. Summary • Statistical modeling is a key approach for quantitative analysis and prediction using network data • Existing statistical network modeling techniques are potentially very powerful • But are currently computationally limited to small networks • Leverage ideas from computer science to extend the “reach” of statistical network modeling to larger networks • Benefits • Computationally tractable modeling of much larger networks • More sophisticated representations for network models • New applications of statistical network modeling

  36. Agenda for MURI Kickoff Meeting at UC Irvine November 18th 2008Location: UC Irvine, Bren Hall, Room 4011 MORNING SESSION8:30  Arrive, continental breakfast available9:00  Introductions and overview of MURI proposal         Padhraic Smyth (UCI)9:30  Research and application challenges from ONR's perspective          Martin Kruger (ONR, MURI Program Manager)10:00 Brief discussion/Q&A session between ONR representatives and PIs

More Related