10 April 2000
Not much to excavate from my brain right now, but I thought I should keep
track of the talks I go hear. So I'm going to start with one I went to
Friday and work backwards.
Andrew Dienstfrey :
Integral Representations of Lattice Sums
Well, pretty much what it says. One gets a certain sum on a lattics as
the solution for particular PDEs with point sources, and one can take
Taylor (or other) series expansions for certain terms that can be switched
with certain integrals... It basically involves switching limits. Big
thing in applied math.
Remember - everything is integration by parts and the Cauchy-Schwartz
inequality. And the regularity of the Lebesgue integral.
Stefano Fusi:
Networks of Integrate & Fire neurons (no leak) - VLSI
Well, there was no leak in his neurons, but there sure was noise. And the
neurons weren't allowed to go below a minimum voltage (a hard
barrier... they bounced). Since there's inherent randomness one can look
at the mean field approximation for firing... and the variance, etc. And
he hooked up a bunch of these suckers together (and cut out the noise? in
synaptic transmission) and looked at steady-state firing rates. Kind of
interesting what happens when it's in a noisy regime.
I still don't think I understand why all these people want to make a chip
to simulate "real" neural networks. Unless one simply wants to understand
what the brain actually does, I don't think it's necessarily the most
efficient way to make a learning system.
Dan Tranchina:
I can't remember the title of his talk (I deleted the email announcement),
but it was about using a simple set of equations (or relatively simple) to
make an approximation of the calcium currents and membrane potential of
retinal cells - the rods and the 2nd order cells connected to them.
Brian Wandell:
Computational Neuroimaging: Color, Motion and Plasticity
I'm not sure I got a bit of the plasticity bit (did he even talk about
that?) but the main thrust was that color-information (aka the signals
from the cones in the retina) was used in V1, the primary visual cortex,
and MT+, a part of the brain associated with motion. Though there is some
tuning related to color and motion, dealing with things of different color
seem to be moving at different speeds even when they're not, one wonders
if the color signal in these areas is really that significant. In any
case, it didn't seem like people were particularly impressed with his
conclusions.
(Brr I'm freezing)
And, apropos of nothing, I've been rooting around on the web because Brian
has gotten me hooked to urbanlegends.about.com and I stumbled upon this
lovely phrase:
Fresh meat is stored primarily in a self-propelled biounit known as a
"pig," which is only slaughtered for major occasions.
From Cecil Adams, the Straight Dope Guy, who then goes onto explain that
this could be a factor in the popularity of Spam among Polynesians.
Oh yeah, it snowed yesterday, and Stu's friend Cindy is here and I met one
of her sister's in-laws, whose husband used to be a deacon, and I sent a
couple pics of myself to Salon.com, and I saw a neat exhibit on Walker
Evans and I really must say I like seeing the quilts at the Metropolitan
Museum. There's this really cool crazy quilt made in mourning for a young
woman named Grace. I want to learn how to crazy quilt. I accidentally
got stuck with a bag of someone else's clothes (who I know will never show
up again (how do I diagram that?)), and I think it would be perfect to cut
them up into scraps and make some crazy quilts. They're colored really
brightly.
And we had an art exhibit here at Courant last Thursday & Friday. You
know I'll always show up for an open bar.