Computer Science Department

Dan Grecu's Homepage

Research area

I am interested in agent learning in complex and changing environments and in multi-agent learning paradigms. More generally, I am interested in investigating in creating agent models about how to learn. Some of the questions on which I have recently focused in my research are:

What triggers a learning process?

Where can the information for learning be found?

What are the factors that control and maintain a learning process?

Can social interaction schemes boost the learning process?

How can agent use learning in complex environments with overwhelming amount of information?

To explore these and other related questions I have pursued two directions of research:

Flexible agent learning in multi-agent design systems

Design is an extremely challenging area for AI. It provides a wide variety of domains and complexity levels to test AI systems. As a human activity which is performed most of the time in teams, it offers a rich collection of examples and paradigms for multi-agent systems research. At WPI we have developed the Single Function Agent model, which we have used to test some of our theories related to multi-agent design.

Multi-agent design provides a serious testbed for agent performance. The size of the problems and the variety of agents involved leaves little hope that an agent has complete knowledge about what happens around it, whether it is the design problem or the agent environment. However, awareness of the events and processes that happen in the multi-agent system can make a considerable difference for an agent's decisions.

In my thesis research I am investigating the possibility of agents coping with highly dynamic environments by using flexible learning mechanisms. Through flexible learning agents try to reconcile contradictions between their beliefs about the environment in which they find themselves and the events in that environment. Flexible learning is desirable in cases where the learning goals arise dynamically, and cannot be decided at the time when agents are implemented. Agents decide autonomously that they need to start a learning process, they reason about where they might find the necessary information to support the learning, and they also have criteria to stop a learning process. Agents are thus continuously revising the knowledge that defines their relationship with the environment in which they act.

Coactive learning

Coactive learning is a multi-agent learning paradigm that we have developed at WPI to investigate collaborative learning schemes. The coacting approach models collaborative learners focusing on the same task that support each other in their learning processes. The learners are assumed to be incremental and they provide each other with feedback and/or support during learning. For example, an agent may consult another agent about how to classify a given training instance, or ask the other agent what knowledge (e.g., rule or training instance) it would use to classify a new instance.

Coactive learning has been successfully used to achieve an emergent noise filtering effect between learners that do not possess noise filtering algorithms, to reduce the number of instances needed by an agent for training, to parallelize learning processes, and to compensate for class distribution biases in the training sets of the agents. The coactive learning model allows to run simulations that combine learners in various interaction schemes, and provides insight into the impact of interaction structure and density on learning.

AI research in our department

Most of the AI research in our department is developed in the

AI Research Group

AI in Design Group

To see what other people are doing in the areas of AI in design and multi-agent development you can use our webliographies:

AI in Design Webliography

Multi-Agent Systems Webliography

My Ph.D. committee

To see who is watching over my shoulder, making sure I get things done, here's my Ph.D. committee:

Lee Becker - Computer Science Department, WPI

David Brown - advisor, Computer Science Department, WPI

Diana Gordon - external member, Navy Center for Applied Research in Artificial Intelligence

Stanley Selkow - Computer Science Department, WPI

Publications

You can find here a list of my recent papers. There is also a Technical Report List of our AI in Research Group, which includes other papers of related interest.

Web sites related to my interests

These are some good starting pointers to collections of resources related to

Teaching

I am teaching two courses:

CS1005 - Introduction to Programming in C

CS2005 - Techniques of Programming

When it's that time of the year you can access from here all the course material. If you need more detailed information about the material to be covered and about the course organization for the coming terms, feel free to ask.

Dan Grecu
Computer Science Department
Worcester Polytechnic Institute
100 Institute Rd.
Worcester, MA 01609, U.S.A.

Email: dgrecu@cs.wpi.edu
Phone: (508) 831-5006
Fax: (508) 831-5776
URL: http://www.cs.wpi.edu/~dgrecu/

Dan Grecu's Homepage

Research area

Publications

Interests

AI in Design Machine Learning

Agents General AI

Psychology

Teaching

Research area

Flexible agent learning in multi-agent design systems

Coactive learning

AI research in our department

My Ph.D. committee

Publications

Web sites related to my interests

AI in Design

Machine Learning

Multi-Agent Systems

General AI

Psychology

Teaching

Last modified: December 11, 1998