Opening Words of the Chair.
As a chair of the Third Oregon Symposium on Logic, Design and Learning,
I welcome everybody very cordially and wish us all true successes in our
presentations and constructive progress in discussions.
This symposium is an outgrowth of work of Prof. Perkowski's group at
PSU, as well as their collaborators outside PSU, on applying logic
synthesis methods to the areas outside traditional circuit design.
We believe that the methods developed in design automation are powerful
and can be successfully used in a number of areas outside logic
synthesis. EDA researchers developed efficient data structures such as
various decision diagrams, general synthesis paradigms such as
functional decomposition, useful concepts such as relational
descriptions of systems; and fundamental algorithms such as
satisfiability.
Although there is a one-to-one mapping between Logic Synthesis and
Constructive Induction methods of Machine Learning; although there are
strong links between Logic Synthesis and Rough Set Theory, the
researchers in these areas are rarely aware that they talk about the
same things using different technical terminology. We believe that
strong synergy between these fields is possible. Logic Synthesis can
also learn from other areas. For example, information theory methods
used for years in Decision Tree Analysis and Machine Learning can be
used in Functional Decomposition.
When we take a basic textbook in pattern recognition, robotics, or
neural networks, we are surprised how much of the material is based on
the same problems as logic synthesis: graph coloring, set covering,
cliques, search, satisfiability. On the other hand, it seems that the
algorithms developed by researchers in logic are more mature and data
structures they use is more sophisticated than in the neighboring areas.
For instance, BDDs are now being used in a number of areas such as
theorem proving, while model checking methods developed in formal
verification of hardware can be applied to robotics. As another example,
covering problems that are solved in robotics have a limited size due to
the use of obsolete algorithms. Now, with the introduction of implicit
algorithms to solve unate and binate covering, we can handle much larger
task encountered in coordinating robot movements.
The goal of the symposium is to investigate these and other mutual links
between logic, design, and learning. Logic is a fundament of design and
learning and its role in many practical design areas is crucial, as
exemplified by the proliferation of logic programming and constraint
programming languages. Recent research shows that logical methods have
their place in learning, and the importance of learning methods in
design also becomes visible, as evidenced by the emergence of the
recursive learning paradigm in testing.
We believe that this synergy of logic, design, and learning is more
general than only logic synthesis for electric circuits, and that it is
applicable to multi-valued and fuzzy networks, arithmetic logic and
other systems used in control, robotics, and image processing. For
example, using Walsh transform in logic design, image processing and
genetic algorithms, uniform decomposition methods for binary,
multi-valued, and fuzzy functions and relations, or arithmetic decision
diagrams in Artificial Intelligence, can be only a one out of many
possible applications.
The other issue, which we want to discuss in this symposium, is the
relation between
university and industrial research. Oregon quickly becomes the Mecca of
design automation industry with companies such as Mentor, Synopsys,
Cadence, and Intel but there has been no local forum to discuss
theoretical problems of applied research. On the other hand, academic
researchers, permanently in need of problems to solve, may turn to
problems that are not practically useful.
For instance, does anybody in industry use Spectral Methods for
synthesis and analysis of switching functions? In the past, the question
was often asked, "Does anybody in industry use EXOR logic"? Now we know
that the answer is Yes, and many would like to know, which university
research areas are of primary interest to industry and why.
Many difficult problems in industry can be solved by academicians with
good mathematics and algorithm background. At this Symposium, we try to
have an honest and tough discussion - how can these two groups, industry
and academia, help one another? What are the most difficult issues in
industry, the toughest nuts to crack? To this purpose, we invited
several researchers from leading companies to help us answer these and
other questions.
Next, in the recent years, we observe the growing influence of
biologically inspired ideas in the areas of logic, design, and learning:
set logic, fuzzy logic, evolutionary computing, artificial neural nets,
to name only a few. Because logic methods influence Data Mining,
Knowledge Discovery in Data Bases, Pattern Recognition and Machine
Learning, they will undoubtedly influence the area of robotics. And here
they will intersect with biologically-motivated design and learning. The
area of Intelligent Robotics, that will hopefully grow in Portland, will
have to incorporate these biological and logical sources; the Electric
Horse project to build a robotic horse that will participate in parade
at Rose Festival is the first example of these efforts.
Finally, we believe in a new kind of conference. All papers have been
carefully reviewed and put on the webpage of the Symposium. The remarks
and opinions have been added to them, so that everybody could be
familiar with the material before the conference. The printed
Proceedings are published in the traditional manner. Standard sessions,
panel sessions, tutorials, and exhibitions are organized with the goals
of achieving open tasks, such as building a "robotic dog" for
high-schools, or a robotic image processing software, rather than only
discussing incremental research problems.
We would like to encourage the participants to share their software and
robotic kits, so that every second year that this symposium is
organized, we could learn more about Logic, Design, and Learning, and
grow towards the same goals as a community. We hope that the result will
be common research projects and publications, joint efforts to write
software and build robots.
Alan Mishchenko
Chair, LDL 2000