Research activity in 1994
During 1994, research of the Machine Learning group in Torino has been
twofold. On the one hand, new learning techniques and strategies have been
studied to face robotics problems. On the other hand, new learning
methodologies have been embedded in learning systems developed in the previous
years by the same group. In particular, some of the researchers have integrated
a learning system with deep reasoning mechanisms, with a special care
towards the examples used in the learning phase.
Other researchers studied the use of genetic algorithms as a
tool to search a hypothesis space, in place of traditional mechanisms. In this
way, results obtained in the previous years have been extended by enlarging the
hypothesis space language and by implementing distributed genetic searching
strategies.
A new line of research has been spotted to learn control functions in the
robotics domain. In particular, the integration of symbolic and numeric
methodologies are under investigation, in order to synthesize fuzzy
controllers.
The above researches have been developed within national and international
research projects, funded by the National Research Council (CNR) and by the
Europeean Community. In particular the B-LEARN2 project, funded by the EC,
Machine Learning techniques have been applied to advanced robotics problems,
with very interesting results.
Results of the above activities have been published in the major journals and
international conferences.
Another group, in connection with prof. Francesco Bergadano (University of
Catania and Messina), has been involved in an emerging area at the intersection
of Machine Learning and Logic Programming: Inductive Logic Programming (ILP),
the induction of Logic Programs from positive and negative examples of their
input output behavior. This field is becoming more and more important, and a
specific ESPRIT project (funded by EC) is expecially devoted to it. The
University of Torino is a main partner in this project, and participates in the
the research with papers published in the main international journals and
conferences, and with the development of learning systems specifically devoted
to the synthesis of logic programs. In Torino, the research is particularly
devoted to Software Engineering applications of ILP. ILP techniques have been
used to assist and automatize software development, testing, maintenance and
reuse. The group is also interested in applications to knowledge discovery in
databases. The results of the research within ILP at the University of Torino
will be published in a book by MIT press available in the second half of 1995.