The next meeting of the SICSA AI Theme will take place on Wednesday, August 30, 2017 at the Cottrell Building Room B3, University of Stirling.
You can register for this event here: https://www.eventbrite.co.uk/e/artificial-intelligence-research-theme-meeting-tickets-36141047847
How do the many facets of learning in AI relate to each other?
Many AI systems learn: the alternative (that the system arrives fully functional, just needing plugged into place) implies the development of systems of staggering complexity if they are to be deployed in the real world. But the adaptation or learning of systems takes many different forms.
- In the neural network arena, learning is associated with changing the strength of links between neurons (and sometimes with introduction and removal of neurons as well);
- it is about robots learning about their sensory systems, and learning about their environments;
- in Bayesian systems, learning is statistical, calculating the probability of hypotheses
- with support vector machines, learning is about learning a decision surface;
- in decision trees, it is about inducing an appropriate tree from examples;
- and there are many other types of learning in AI in other branches of the discipline.
There is growing interest in this area, as evidenced by the DARPA call “Toward Machines that Improve with Experience”, see http://www.darpa.mil/news-events/2017-03-16,
The aim of this meeting is to seek connections between these different types of learning. At first sight many of them seem quite independent of each other, but are they?
Ian Murray, School of Informatics, University of Edinburgh.
David King, Computing and Mathematics, Abertay University
(Other invited speakers to be advised
We intend to have several other speakers as well as a poster session: we seek abstracts both for speakers and for poster presentations from both established researchers and PhD students.
Abstracts should be 2 pages or less, and should be on learning in different aspects of AI, both theoretical and applied, (NNs, SVMs, robotics, gaming, blackboard systems, etc.). Please state whether you would prefer a poster, a short talk, or would be happy with either.
These should be sent to Professor Leslie Smith (mailto:firstname.lastname@example.org), as a .pdf file, before 31 July 2017: submissions will be acknowledged.
Meeting Format: The meeting will start at 10.15 with coffee on arrival: there will be talks from the invited speakers, contributed talks, and a poster session.
We intend to organize some discussion groups in the afternoon to investigate possible areas of collaboration. The aim is to form some conclusions about novel relationships between learning in different aspects of AI as a result.
The meeting will end about 16.00, probably adjourning to less formal premises to continue discussions.