An Integrated Approach to Decision Making under Uncertainty UCLA: A. Darwiche, P. Kellman, J. Pearl

1
An Integrated Approach to Decision Making under
UncertaintyUCLA A. Darwiche, P. Kellman, J.
PearlUCI R. Dechter, S. IraniUIUC D. Roth
2
Project Objectives

• Develop basic methods for helping human-decision
  makers attain their full potential in uncertain
  environments
• Integrate the developed methods to illustrate
  their utility in enhancing the decision making
  process (two levels of integration)

3
Requires Integration of Techniques for..

• Representing uncertain, incomplete information
• Fusing uncertain information of different kinds
• Evaluating and ranking courses of actions
• Providing real-time incremental responses to
  user queries
• Interacting with users in cognitively grounded
  manner

4
Elements of Research Program

• Probability theory
• Knowledge representation and reasoning
• Algorithms
• Natural language processing
• Machine learning
• Cognitive Science/Psychology

5
Key Commitments

• Probability theory as the foundation for managing
  uncertain information
• Bayesian belief networks as the mechanism for
  realizing computer implementations of uncertainty
  methods
• Mix of planned theoretical developments and
  practical implementations

6
(No Transcript)
7
AFRL Dynamic Command and Control Branch

• Using Bayesian techniques for temporal
  propagation and analysis of the causal
  uncertainties involved in military plans, where
• Effects are often delayed and will have only
  finite persistence
• Introduction and analysis of time-based evidence
  is commonplace
• Dynamic prediction of value of information is
  critical.
• Group has been analyzing national anti-terrorism
  plans using Bayesian networks
• Network built by real planners at the top level
  in DOD
• Briefed at the JCS (Joint Chiefs of Staff) level
  and to the State Department

8
(No Transcript)
9
(No Transcript)
10
AFRL Dynamic Command and Control

• Technology considerations
• Building models
• It is meaningless to construct models from
  historical data
• Parameter estimation is a major concern
• Causal structures constructed by experts tend to
  fail as I-maps.
• Basic reasoning with models (probabilities of
  events)
• Useful models contain on the order of a thousand
  nodes
• Timely computation of answers to meaningful user
  questions is critical
• Sophisticated reasoning with models
  (non-existent)
• Praeto optimal plans (cost effective plans that
  meet a threshold of success)
• Meaningful display of results
• What other useful information can we obtain from
  a model?

11
Schedule

• 0900-0915 Opening remarks
• 0915-1000 Darwiche
• 1000-1055 Pearl, Tian, Brito
• 1055-1110 Break
• 1110-1145 Hopkins, Bonet
• 1145-1230 Kellman
• 1230-1400 Lunch
• 1400-1500 Dechter, Bidyuk
• 1500-1530 Irani
• 1530-1545 Break
• 1545-1615 Darwiche
• 1615-1715 Roth

12
UCLA

• Applied Logic Bayesianism and Causality, or, Why
  I am Only a Half-Bayesian,Pearl
• UAI-02 Qualitative MDPs and POMDPs An
  Order-of-Magnitude approximationBonet and Pearl
• UAI-02 Generalized Instrumental VariablesBrito
  and Pearl
• UAI-02 On the Testable Implications of Causal
  Models with Hidden VariablesTian and Pearl,.
• AAAI-02A Graphical Criterion for the
  Identification of Causal Effects in Linear
  ModelsBrito and Pearl
• AAAI-02 Strategies for Determining Causes of
  EventsHopkins
• AAAI-02 A New Characterization of the
  Experimental Implications of Causal Bayesian
  NetworksTian and Pearl
• AAAI-02 A General Identification Condition for
  Causal EffectsTian and Pearl
• R-301 Causality and Counterfactuals in the
  Situation CalculusHopkins and Pearl

13
UCLA

• KR02 A logical approach to factoring belief
  networks Adnan Darwiche
• AAAI02 A distance measure for bounding
  probabilistic belief change Hei Chan
  and Adnan Darwiche
• AAAI02 A compiler for deterministic
  decomposable negation normal form
  Adnan Darwiche
• AAAI02 Using weighted MAX-SAT to approximate
  MPE James Park
• UAI02 MAP complexity results and
  approximation methods James Park
• D-118 A differential semantics for jointree
  algorithms James Park and Adnan
  Darwiche
• D-130 Optimal time-space tradeoffs in
  probabilistic inference David Allen and
  Adnan Darwiche

14
UCLA
Elsevier Science Press. From fragments to
objects Segmentation and grouping in vision.
Shipley Kellman
Encyclopedia of Cognitive Science.
Vision-occlusion, illusory contours and filling
in." Kellman
Experimental Child Psychology. Separating
processes in object perception. Kellman
Vision Research. Surface integration influences
depth discrimination.Kellman Shipley
Symposium on cognition. Segmentation and grouping
in object perception A 4- dimensional
approach. Kellman
Stevens' handbook of experimental psychology.
Perceptual learning.Kellman
15
UCI

• AAAI-02 Mini-clustering a bounded anytime
  inferenceMateescu, Dechter and Kask
• UAI-02 Iterative join-graph propagation
  Dechter, Kask and Mateescu
• UCI-02 Cutset sampling Bidyuk and Dechter
• UCI-02 Reasoning with partially deterministic
  information Larkin and Dechter
• AAAI-02 Generating random scenarios with
  constraints Dechter, Kask, Emek and Bin
• UCI-02 Epsilon-cutset or When iterative belief
  propagation works? Bidyuk and Dechter

16
UCI

• An Asymptotically Optimal Algorithm for the
  Dynamic Traveling Repair Problem. Xiangwen Lu,
  Amelia Regan and Sandy Irani.Proceedings of the
  Transportation Research Board Annual Meeting,
  2002.
• Scheduling with Conflicts on Bipartite and
  Interval Graphs. Sandy Irani, Vitus Leung.
  Journal of Scheduling Special Issue on Online
  Algorithms.
• Competitive Analysis of Dynamic Power Management
  Strategies for Systems with Multiple Power
  Savings States. Sandy Irani, Sandeep Shukla,
  Rajesh Gupta. Design Automation and Test In
  Europe, 2002.
• On-Line Algorithms for the Dynamic Traveling
  Repair Problem. Sandy Irani, Xiangwen Lu, Amelia
  Regan. Symposium on Discrete Algorithms, 2002.
• An Analysis of System Level Power Management
  Algorithms and Their Effects on Latency. Dinesh
  Ramanathan, Sandy Irani, Rajesh Gupta. IEEE
  Transactions on CAD.

17
UIUC

• COLING02 Probabilistic Reasoning for Entity and
  Relation Recognition
• D. Roth and W-T. Yih
• COLING02 Learning Question Classifiers
• X. Li, and D. Roth
• ICML'02 On generalization bounds, projection
  profile, and margin distribution
• A. Garg, S. Har-Peled and D. Roth
• ECML'02 Learning and Inference for Clause
  Identification
• X. Carreras, L. M\arquez, V. Punyakanok and D.
  Roth
• ILP'02 Learning with Feature Description Logics
• C. Cumby and D. Roth
• UIUCDCS-R-2002 Constraint Classification A New
  Approach to Multiclass Classification
• S. Har-Peled and D. Roth and D. Zima