AI, Machine Learning, Algorithms

AI group

Overview

This research cluster works to develop technology, processes, and software to enable effective access to and utilization of overwhelming amounts of information. The IIS cluster works to combine knowledge from database, machine learning, information retrieval, networking and human-computer interaction research to create more intelligent information systems. Core strengths include collaborative filtering, probabilistic modeling, spatial databases, usability engineering, web-based interfaces, and wireless computing.

We seek to construct computer systems that can build models of their environments and apply those models to make reliable, rapid decisions. We develop new methods for statistical learning, data mining, and probabilistic reasoning and apply these to problems in environmental monitoring, ecological science, manufacturing engineering, space exploration, robot control, and web-based information systems. Anticipated impacts include cheaper and more accurate environmental monitoring, more cost-efficient factories, and easier access to information on the web.

Research Thrusts

AI/Machine Learning

Decision-Making & Reinforcement Learning
Machine Learning & Data Mining
Pattern Recognition
Probabilistic Representation & Reasoning

Algorithms

Algorithms
Approximation Algorithms
Graph Algorithms
Computational Geometry

Related Courses

CS 515: Algorithms and Data Structures
CS 517: Theory of Computation
CS 523: Advanced Algorithms
CS 531: Artificial Intelligence
CS 533: Advanced Artificial Intelligence
CS 534: Machine Learning
CS 536: Introduction to Graphical Models
CS 539: Special Topics in Artificial Intelligence

Software Downloads

StratagusAI: an open-source modification of the Stratagus RTS game engine to support AI research
Java Library for Adaptation-Based Programming
Error Correcting Output Codes for multiclass learning problems
MAXQ Hierarchical Reinforcement Learning system
TreeCRF system for training conditional random fields via functional gradient ascent
PCBR region detector
Open Source SIFT Feature Detector
TaskTracer system

Faculty

Glencora Borradaile
Algorthms; approximation algorithms; computational geometry; planar graph algorithms; graph algorithms

Alan Fern
Artificial intelligence, including machine learning, data mining, and automated planning/control

Raviv Raich
Adaptive sensing/sampling; manifold learning; sparse representations for signal processing

Weng-Keen Wong
Data mining; machine learning; anomaly detection; human-in-the-loop learning; ecosystem informatics

Tom Dietterich
Machine learning; intelligent systems; intelligent user interfaces; ecosystem informatics

Xiaoli Fern
Machine learning and data mining, specifically the subfields of unsupervised learning and pattern discovery

Prasad Tadepalli
Artificial intelligence; machine learning; automated planning; natural language processing

Research Facilities

AI LaboratoryAI Computer Cluster: 34 12-core 64-bit processors (408 total cores)

Research Partners

Industry & Other Organizations

Universities

Selected Projects

Adaptation-Based Programming
(A. Fern, M. Erwig, T. Nguyen)

Integration of programming language and machine learning research
Allow programmers to leave difficult program choices as open choice points
Use machine learning techniques to automatically optimize choice points
Applications to software game agents and network optimization

Automated Planning and Learning for Complex Environments
(A. Fern, P. Tadepalli)

Development of learning and planning algorithms for selecting actions in complex environments
New monte-carlo planning algorithms
New algorithms for reinforcement learning
Applications to emergency response planning, game-playing agents, ecological conservation

Understanding Visual Scenes and Activity
(T. Dietterich, A. Fern, S. Todorovic)

BugID project: recognizing bug species from image data
The OSU Digital Scout Project: Computer Vision Meets American Football
Generic part-based object recognition and learning
Generic high-level event recognition and learning

Ecosystem Informatics
(T. Dietterich, X. Fern, W-K. Wong)

Learning models of species distribution, dispersal, and migration
Occupancy-detection-expertise models to understand both the distribution of species and the process of observing them
Recognition of bird species from songs and flight calls
Optimal active management of wildfires and invasive species

End-User Training of Machine Learning Systems
(M. Burnett, T. Dietterich, A. Fern, W-K. Wong)

Enable end-users to tune adaptive software systems
Investigating effective forms of communication between computer students and human teachers
Applications to labeling and ranking systems
Applications to teaching autonomous control agents

Learning from Natural Language Texts
(T. Dietterich, X. Fern, P. Tadepalli)

Learning general rules by reading text
Learning from incomplete and noisy data
Incorporating pragmatics of document generation in learning from texts

Intelligent User Interfaces: The TaskTracer Project
(T. Dietterich)

Windows add-on to support multi-tasking desktop workers
Methods for machine learning tagging of email, documents, and web pages
User interface support for information re-finding and interruption recovery

Genome Informatics
(W-K. Wong, T.C. Mockler)

Improved methods and tools for de novo assembly of high throughput sequencing data.
New methods and tools for the automatically accurate annotation of genomes in near real-time.
Design and implementation of novel tools for interacting with massive biological datasets.

Active Transfer Learning
A. Fern, P. Tadepalli