 The AI research group is a growing community with a large number of current projects as well as past successes. The group works closely with the Database Systems, Information Retrieval, and Intelligent Internet Systems group and the Computer Graphics, Computer Vision and Animation group.
New projects!![[current]](http://www.cs.washington.edu/ai/green-bullet.jpg) Sensing and Modeling Dynamic Social Networks
ARNAULD: Preference Elicitation For Interface Optimization
Markov Logic Networks: Probabilistic first-order knowledge bases
Planning for Concurrent Durative Uncertain Actions: Augmenting Markov Decision Processes to handle concurrent temporally-extended actions.
Spectral graph partitioning: Clustering and learning in networks of symmetric and asymmetric relationships.
KnowItAll: Automated, domain-independent, web-scale information extraction and evaluation.
Statistical Relational Learning: Learning from noisy data in rich representations.
Collective Knowledge Bases: Merging knowledge from a multitude of sources.
Large-Scale Machine Learning: Mining massive data streams.
Assisted Cognition: Computer systems to aid people with Alzheimer's disease.
CORE: Optimizing search algorithms using Bayesian models to predict running time.
In the list below current projects are marked with a while past work is noted with a  . KnowItAll: Automated, domain-independent, web-scale information extraction and evaluation.
Tukwila: Data integration system for heterogeneous data on the web.
Mulder: A natural-language question-answering service that Believes.
Tiramisu: Declarative Web-site management.
The Internet Softbot: The mother of all intelligent internet systems.
Site popularity meta-search: Re-ranking results of web engines using web page popularity.
Grouper: Document clustering for improved search results on the web.
ReferralWeb: Explore the social networks that exist on the Web.
Intelligent and Personalizable User Interfaces ARNAULD: Preference Elicitation For Interface Optimization
SUPPLE: Automatic Generation of User Interfaces
Provably Reliable Question-Answering Interfaces: Natural language interfaces that are guaranteed to answer "easy questions" correctly.
Adaptive Web sites: Sites that improve their organization by learning from visitor usage.
Web Site Personalizers: Intermediaries between servers and visitors that automatically adapt and customize content for wireless web visitors.
Programming by Demonstration: Using AI techniques to improve user interfaces.
Adaptive interfaces for machine learning systems.
Knowledge Representation and Reasoning CORE: Optimizing search algorithms using Bayesian models to predict running time.
Structural Modeling for Anatomy: Representing knowledge about human anatomy.
Walksat: Stochastic local search for satisfiability.
Sensing and Modeling Dynamic Social Networks
Activity Recognition
Assisted Cognition: Computer systems to aid people with Alzheimer's Disease.
Markov Logic Networks: Probabilistic first-order knowledge bases
Statistical Relational Learning: Learning from noisy data in rich representations
Collective Knowledge Bases: Merging knowledge from a multitude of sources
Large-Scale Machine Learning: Mining massive data streams
Belief networks.
Spectral graph partitioning: Clustering and learning in networks of symmetric and asymmetric relationships
Statistical machine learning.
LSD: Learning source descriptions for data integration.
CMM: Converting model ensembles into a single comprehensible model.
RISE: High-performance concept learner, unifies rule induction and instance-based learning.
Naive Bayes.
MetaCost: Making error-based learners cost-sensitive.
Process-Oriented Evaluation: Avoiding overfitting by estimating a hypothesis' generalization error as a function of the search process that led to it.
3D object recognition.
Content-based image retrieval.
Probabilistic models of the brain.
Spike-based computing and learning: For instance, Temporal sequence learning.
Learning algorithms for vision: For instance, Invariant coding under image transformations.
Brain-computer interfaces: EEG-based systems that allow completely paralyzed patients to interact with a computer.
Planning for Concurrent Durative Uncertain Actions: Augmenting Markov Decision Processes to handle concurrent temporally-extended actions.
Interleaved Contingent Planning and Execution An architecture, motivated by NASA applications, which unifies contingent planning and reactive execution.
blackbox: A planning system that combines SAT technology with Graphplan.
TGP: A an extremely fast temporal planner, aimed at NASA spacecraft domains.
SGP: Handles uncertainty, sensory actions and combines conformant and contingent planning.
LPSAT: Combining linear programming with satisfiability.
Markov Decision Processes
Medic: Compiles STRIPS problems into satisfiability problems. (IJCAI-97 paper)
PYRRHUS: Finds optimal plans for goal-directed value functions.
ZENO: Temporal planner handles deadline goals and continuous change.
BURIDAN: Probabilistic planner reasons about uncertainty.
CBURIDAN: Extends BURIDAN with sensing actions and contingent execution.
XII: Executes sensing actions to handle incomplete information with the Internet softbot.
Occam: Planner optimized for information gathering and controls the Razor softbot.
UCPOP: (distributed to 100+ sites) Handles universal quantification and conditional effects.
FABIAN: Plans with abstract actions which it automatically generates.
Monte Carlo Localization (MCL): Particle filters for state estimation in mobile robotics.
Multirobot systems: Navigation and coordination of multiple robots.
Mobile robot control: Probabilistic techniques that can handle position and sensor uncertainty.
Museum tour-guides: Rhino and Minerva guide visitors through crowded museums.
Assisted Cognition: Computer systems to aid people with Alzheimer's Disease.
| 
CSE Home
