Value-centered Information Theory for Adaptive Learning, Inference, Tracking, and Exploitation
Program Manager: Dr. Liyi Dai, ARO
- University of Michigan: Alfred Hero (PI), Raj Rao Nadakuditi
- MIT: John Fisher, Jonathan How, Alan Willsky
- Ohio State University: Randolph Moses, Emre Ertin
- University of California, Berkeley: Michael Jordan
- University of California, Los Angeles: Stefano Soatto
- University of California, San Diego: Angela Yu
- Arizona State University: Doug Cochran
A complete roster and web page links may be found at People.
Our research program aims to lay the foundations for a new systems theory that applies to general controlled information gathering and inference systems with mission planning. Our research approach comprises three inter-related research themes that collectively address the most critical research challenges. These thrusts are:
- information-driven structure learning and representation
- distributed information fusion
- active resource management for effective information exploitation.
We aim to develop an end-to-end framework that will result in better raw sensor data acquisition and processing, improved fusion of multiple sources and modalities, and more effective sensor management and control.
An executive summary of the project can be found here.
Third-year review meeting
Time and Location
Thursday, Nov 20th, 2014
8:30 am - 4:30 pm
AMP Lab, Soda Hall
Time Activity 8:00 – 8:30 Get settled with coffee 8:30 - 8:35 Welcome (Liyi Dai) 8:35 - 8:50 Project overview (Al Hero) 8:50 - 9:30 Mathematical foundations for value-of-information (Al Hero and Doug Cochran) 9:30 - 10:10 Learning with information constraints (Michael Jordan) 10:10 - 10:30 Break 10:30 - 11:30 Poster highlight talks 11:30 - 1:30 Poster session and lunch 1:30 - 2:10 Information fusion and exploration (Stefano Soatto) 2:10 - 2:50 Information planning and active data collection (John Fisher and Jon How) 2:50 - 3:10 Break 3:10 - 3:30 Radar testbed (Emre Ertin) 3:30 - 3:45 Wrapup (Al Hero) 3:45 - 4:30 Government discussion and de-briefing 4:30 Adjourn
Slides for third-year review meeting
- Project overview (Hero): 1Hero_Overview_Year3_v1.pdf
- Mathematical foundations for value-of-information (Hero & Cochran): 2Hero_Foundations_Year3.pdf
- Learning with information constraints (Jordan): 3Jordan_Learning_Year3.pdf
- Information fusion and exploration (Soatto): 4Soatto_Fusion-Exploration_Year3.pdf
- Information planning and active data collection, Part I (How): 5aHow_AdaptiveSearch_Year3.pdf
- Information planning and active data collection, Part II (Fisher) 5bFisher_Information_Planning_Year3.pdf
- Radar testbed (Ertin): 6Ertin_RadarTestbed_Year3.pdf
- Wrap-up (Hero): 7Hero_Wrapup_Year3.pdf
Poster titles and presenters
- Gene T. Whipps, Emre Ertin, Randolph L. Moses, "A consensus based decentralized EM for a mixture of factor analyzers“ WhippsErtinMoses_poster_overview.pdf
- Diyan Teng and Emre Ertin, “Learning to Aggregate Information for Sequential Inferences” TengErtin_poster_overview.pdf
- Nithin Sugavanam, Emre Ertin, “Optimal measurement matrix design for structured signals in clutter and noise”
- K. Greenewald, A.O. Hero, "Regularized Block Toeplitz Covariance Matrix Estimation via Kronecker Product Expansions“ GreenewaldHero_poster_overview.pdf
- H.W. Chung, T. Tsiligkaridis, B. Sadler, A. Hero, “Collaborative 20 questions” ChungHero_poster_overview.pdf
- Z. Meng, B. Erikson, A.O. Hero, "Learning Latent Variable Gaussian Graphical Models" MengHero_poster_overview.pdf
- Zhang, Chen, Zhou and Jordan: Spectral methods meet EM: A provably optimal algorithm for crowdsourcing
- Zhang, Wainwright and Jordan: Lower bounds on the performance of polynomial-time algorithms
- Nishihara, Jegelka and Jordan: On the convergence rate of decomposable submodular function minimization
- J. Duchi, M. Jordan, M. Wainwright and Wibisono: Optimal rates for zero-order optimization: the power of two function evaluations
- Brian E. Moore, Raj Rao Nadakuditi, and Jeffrey A. Fessler, “Improved robust PCA using low-rank denoising with optimal singular value shrinkage”
- Lauren Crider, Douglas Cochran, “Value of Information Sharing in Network Signal Detection.” CriderCochran_poster_overview.pdf
- S. Zhang, A. Tran and A. Yu, “Personality and Behavioral Predictors of Human Exploration/Exploitation Behavior in a Bandit Task”
- B. Mu and J. How, “Focused information gathering with an application in SLAM”
- Josh Hernandez, Konstantine Tsotsos, Gottfried Graber, and Stefano Soatto, “Scene Segmentation with Dense Reconstruction from Monocular Video“
- Konstantine Tsotsos and Stefano Soatto, “Robust Visual-Inertial Navigation and Mapping"
ARL Visit June 14 2014
PDF slide presentation (Hero): ARL_2014_slides.pdf
PDF slide presentation (Yu): ARL14 yu.pdf
Second-year review meeting
Monday, September 9th, 2013
8:30 am - 4:30 pm
Seminar Room, Zahl Building
U.S. Army Research Laboratory
2800 Powder Mill Road, Adelphi MD 20783
ARL visitor pamphlet: .pdf
Non-ARL personnel should arrange invitations for attendance through Dr. Liyi Dai, ARO
Morning of the review:
Visitors budget 30 minutes to get onto base, get badged, and be escorted to the seminar room. Parking at ARL can be challenging, so allowing additional time for this is also advisable. Visitors should try to park in the South Lot. Late arrivals may have to park in K Lot. Visitors must enter through building 205 and report to the badging office. Government issued ID will be required (2 forms of government issued ID required for non-US citizens). This building and the parking lots are labeled 1n the ARL visitors pamphlet. US visitors will be escorted from the badging room to the meeting room in Zahl. Non-US visitors will need to be escorted from the main front gate to parking, badging, and then the meeting room. Visit requests, filed in advance of the meeting, let the front gate guards know the appropriate escorts for the non-US visitors.
Time Activity 8:00 – 8:30 Check in at ARL visitor office 8:30 - 8:35 Welcome, Liyi Dai 8:35 - 8:50 Project overview, Al Hero 8:50 - 10:15 Thrust area I: Learning and Representation Overview, Michael Jordan PI Summaries, Michael Jordan, Stefano Soatto, Al Hero, Angela Yu 10:15 - 10:30 Break 10:30 - 11:35 Thrust area III: Information Exploitation Overview, Doug Cochran PI summaries: Emre Ertin, Jon How, Doug Cochran 11:35 - 12:05 Collaborative highlights and DURIP testbed (Emre Ertin) 12:05 - 1:50 Lunch and poster session 1:50 - 2:55 Thrust area II: Information Fusion Overview, John Fisher PI summaries, Raj Rao Nadakuditi, Randy Moses, John Fisher 2:55 - 3:05 Wrap-up, Al Hero 3:05 - 4:05 Government discussion and de-briefing 4:05 Adjourn
Slides for second-year review meeting
Project overview (Hero): 1Hero_overview_13_v.pdf
Learning thrust area overview (Jordan): Media:Jordan_thrust1_overview.pdf
- Learning thrust subarea 1 (Jordan): Media:3jordan_progress_report.pdf
- Learning thrust subarea 2 (Soatto): Media:4AroMuriReviewSoatto2013.pdf, File:Search demo.mp4, File:Searchdemo2.mp4
- Learning thrust subarea 3 (Hero): 5Hero_thrust1_13.pdf
- Learning thrust subarea 4 (Yu): Media:6yu.pdf
Exploitation thrust area overview (Cochran):
- Exploitation thrust subarea 1 (Ertin): Ertin8MURI2013.pdf
- Exploitation thrust subarea 2 (How): 9How_2013_muri_review.pdf
- Exploitation thrust subarea 3 (Cochran): 10Cochran_thrust3_13.pdf
Fusion thrust area overview (Fisher):
- Fusion thrust subarea 1 (Nadakuditi): 13Nadakuditi_MURI.pdf
- Fusion thrust subarea 2 (Moses): 14Moses_thrust2_13.pdf
- Fusion thrust subarea 3 (Fisher): 15Fisher_thrust2_2013.pdf
Wrap-up (Hero): 16Hero_wrapup_13_v.pdf
Poster titles and presenters
- N. Asendorf and R. R. Nadakuditi, "The Performance of Deterministic Matched Subspace Detectors: Informative Versus Useful Subspace Components"
- T. Broderick, N. Boyd, A. Wibisono, A. Wilson, and M.I. Jordan, "Streaming Variational Bayes"
- L. Crider, R. R. Nadakuditi, and D. Cochran, "Analyze-Fuse vs. Fuse-Analyze in Multiple-channel Emitter Detection"
- Z. Meng,", D. Wei, A. Wiesel, and A. Hero, "Distributed learning of Gaussuab graphical models via marginal likelihoods"
- B. Mu, G. Newstadt, J. How, and A. Hero, "Heterogeneous Agent Target Assignment in an Unknown Environment"
- B. Mu, G. Chowdhary, J. P. How, "VoI Aware Active Task Assignment Hardware Experiment"
- G. Papachristoudis and J. W. Fisher, "Efficient Information Planning in Graphical Models"
- J. Straub and J. W. Fisher, "Information gathering with mobile agents: sensor and model complexity tradeoffs"
- N. Sugavanam, and E. Ertin, "Robust Filtering with Uncertainty in Initial State"
- N. Sugavanam, and E. Ertin, "Sensor Selection Problems in Adversarial Setting"
- R. T. Suryaprakash and R. R. Nadakuditi, "The Performance of MUSIC-Based DOA In White Noise With Missing Data"
- D. Teng, and E. Ertin, "Optimal Quantization of Likelihood for Low Complexity Sequential Testing"
- D. Teng,and E. Ertin, "Learning Statistical Manifolds for Target Signatures"
- T. Tsiligkaridis and A. Hero, "Covariance Estimation In High Dimensions via Kronecker Product Expansions"
- T. Tsiligkaridis, B. Sadler, and A. Hero, "Collaborative 20 Questions for Target Localization"
- G. T. Whipps, E. Ertin, and R. L. Moses "Decentralized EM for a Mixture of Factor Analyzers"
- G. T. Whipps, E. Ertin, and R. L. Moses "Distributed Detection with Collisions in a Single-hop Wireless Sensor Network"
- S. Zhang and A. J. Yu, "Forgetful Bayes and Myopic Planning: Human Decision Making in a Bandit Setting"
- Y. Zeng, C. Wang, S. Soatto, and S.-T. Yau, "Nonlinearly Constrained MRFs: Exploring the Intrinsic Dimensions of Higher-Order Cliques"
Collaborative highlights topics
- Use of the DURIP testbed capabilities (Ertin, Hero, Moses, How, Fisher, Cochran)
- Convex VoI criteria for mission planning (Hero, How)
- Information-geometric VoI in sensor planning (Cochran, Hero)
- Learning how to rank many objects (Jordan, Hero)
- Improving information via human-in-the-loop processing (Hero, Jordan, Yu)
- Non-parametric estimation of Fisher information (Ertin, Hero)
- Quantized Belief-States and Confirmation Bias (Ertin, Yu)
- Kronecker product covariance representations in distributed sensing (Ertin, Hero)
- Analyze-Fuse versus Fuse-Analyze in multi-channel detection (Nadakuditi, Cochran)
First year review meeting
Monday, October 29, 2012
8:00 AM - 4:00 PM
Rice Conference Room, 6764 Boelter Hall
UCLA, Los Angeles, CA 90095
Parking information for those with cars:
For those arriving by car from the South, exit the 405 freeway at Wilshire East, then follow Wilshire East to Westwood North, that end into campus. Past the hospitals there is a parking booth in the center divider. Stop there and get a daily permit for $11. Also ask them for a campus map. You will go to Boelter Hall, that is on your right behind the building that carries the sign "Henry Samueli School of Engineering and Applied Science". We will meet in the Rice Conference Room on the 6th floor, south-west corner of the building. For any problems, call Stefano Soatto at (310) 869 2972.
The formal agenda will start at 8:30AM with overview and presentations from the VoI MURI team. We will break for lunch and posters (presented by MURI-supported students) at around 11:30AM. We will then resume at 1PM with more presentations and wrapup, ending at around 2:30. This will be followed by a government caucus and feedback at around 3:30PM to the MURI team.
Thursday, October 13, 2011
8:30 AM - 4:30 PM
Johnson Rooms, 3rd floor, Lurie Engineering Center
1221 Beal Avenue, Ann Arbor, MI 48109
For visitor information, please click here.
Wednesday, October 12, 2011
10 AM - 5 PM
George Haddad Conference Room, 3rd floor, EECS building
1301 Beal Avenue, Ann Arbor, MI 48109-2122
collaborations: collaborative activities between co-PI's.
publications: links to publications resulting from this program.
reports: links to reports on this program.
code: links to software related to this research.
Please email to obtain a username and password to edit this wiki. All affiliated faculty, researchers, and students are encouraged to have access. We hope to make this the central distribution center for information, files, etc.
To the private section of this wiki.