🚀 go-pugleaf

Subject:
             [Fwd: trip report for IWANN'99]
        Date:
             Thu, 24 Jun 1999 10:48:37 -0600
        From:
             Kathi Nietfeld <nietfeld@CS.ColoState.EDU>
 Organization:
             CSU Computer Science Department
          To:
             faculty@CS.ColoState.EDU, graduate@CS.ColoState.EDU,
staff@CS.ColoState.EDU




 Subject:
         trip report for IWANN'99
   Date:
         Thu, 24 Jun 1999 08:46:20 -0600 (MDT)
   From:
         charles anderson <anderson@CS.ColoState.EDU>
     To:
         nietfeld@CS.ColoState.EDU



>From June 2nd to June 4th I attended the 5th International
Work-Conference on Artificial and Natural Neural Networks in Alicante,
Spain.  Yeah, this was rough.  If you ask, I'll show you pictures of
the beach outside the meeting rooms.  This bi-annual conference is
always held in Spain.  Attendees are from many parts of Europe and
some from the US.  The conference is kind of like a southern Europe
version of the NIPS conference held in Denver every year, in that it
combines neuroscience and artificial neural networks.

I participated in and co-chaired an invited session on the "Analysis
of Brainwaves".  This was organized by Jose Millan, who shares my
interests in brainwaves and in reinforcement learning.  He has been
the key figure in forming a European collaboration of researchers
from Rome to Helsinki to work on brain-computer interfaces.  We hope
to develop a funded collaboration soon.

I also presented two other papers, one by Matt Kretchmar on his ideas
about temporal neighborhoods for function approximation in
reinforcement learning, and one by David Peterson on independent
components analysis for EEG.  Both received a lot of interest from the
audience.  Thanks Matt and Dave.

Most of the talks were of fairly high quality.  A very interesting
session was one on non-linear approaches to independent components
analysis, a twist on the standard principal components analysis.  This
has potential for separating the various sources that combine to
produce EEG signals.

The conference included a super list of invited speakers, including
Poggio from MIT (who had to cancel at the last minute) and Eve Marder
from Boston University.

The most interesting paper on reinforcement learning was by Andres
Perez-Uribe (http://lslwww.epfl.ch/~aperez/Research.html).  He has
placed an FPGA board on a Khepera.  The FPGA implements reinforcement
learning algorithms, including Q-learning and Sutton's DYNA algorithm,
all autonomously.  Impressive!  There were some other interesting
talks on FPGAs.  One person showed a complete VHDL program that
implements the forward and backward phase of error backpropagation.
Another discussed their use of on-line arithmetic, by which they are
able to do multiplication from left to right, so something like a
weighted sum consisting of multiplication followed by addition can be
pipelined.

The organizers treated us very well.  One night we toured the Castle
of Santa Barbara that towers over the Alicante beaches.  The oldest
parts of the castle are about 2,000 years old.  As the sun set, local
dancers and musicians entertained us inside the castle.  The next
night we had the official conference dinner at a local restaurant,
after which the local "tuna" entertained us.  A tuna, I guess, is kind
of like the local university's pep band.  They play mandolins,
guitars, and tamborines, sing school spirit songs and sing for beer.


                Chuck