My research interest lies in understanding and designing intelligent
systems that adapt to changes in their environments. The primary goals
at present are modeling modularized architecture with pattern
recognition and learning abilities, and applying these models to the
design of intelligent interface systems.
Toward these ends, my students and I are working on a broad range of
topics. First, we are formulating a computer model of multi-agent
architecture that has the ability to learn. It is applied to our
working system, Michele (Multi-agent Interface system with
Communication by Hectic ELEments), to support cooperative work by
learning from examples. To enhance the adaptability of our system,
much work is devoted to learning and pattern recognition. This work
has produced applications such as a system that learns symbolic
inference programs from raster-scanned images, a system that
transforms hand-drawn tables into databases, and an intelligent
multimedia document-handling system integrated with hypertext
databases.
By incorporating these into Michele, we intend to construct a
modularized interface system with learning and pattern recognition
abilities for distributed computing environments.
Second, we are designing and implementing a model of modularized
neural computing architecture called $\mu$-net (MUlti-agent neural
NETwork architecture). The model is basically a composite network of
neural-network modules, where each module learns to perform a
particular function that helps the whole network adapt to the
environment. Currently it is being applied to the domain of
autonomous navigation.
We are also committed to research that supports
the construction and implementation of our models. The research
agenda includes distributed operating systems and programming
environments, multimedia databases and networks, constraint
programming languages, efficient search algorithms, and generation of
new representations in problem solving.