IZHIKEVICH Eugene SPIKES 29 Jan 2013
<p class="MsoNormal" >IZHIKEVICH Eugene SPIKES 29 Jan 2013</p>
<p>
he is trying to hire researchers for his lab: </p>
<p>
he has 21 people; trying to hire more ? 19 more...</p>
<p>
2006 he started scholarpedia!</p>
<p>
also began encyc of computational neurosci....</p>
<p>
<note to myself: reread nick steinmetz article on frontal eye fields></p>
<p>
was at NSI neurosci instit with Gerry Edelman</p>
<p>
<b>how do our brains work?... SPIKES!</b></p>
<p>
he believes it : <b>spikes and not rates </b> (but there is no proof)
this is a religion with him!</p>
<p>
spike timing:</p>
<p>
he shows a plot of 100 neurons firing on avg 1 per sec (so all at 1 hz but
it it's spike timing... then it's <b>100 factorial! 100160 (because it's all order dependent...</b>
can convey far more information... and, of course, that is true.)</p>
<p>
<b>spike timing is everything</b>...</p>
<p>
he says 100B neurons and 1 Quad synapses...
it HAS to use spike timing for neural computations!</p>
<p>
PLAN for today: everything is spike timing (NOT rates!!!)</p>
<p>
Gordon Shepard 1981...
spikes are ubiquitous...</p>
<p>
he shows cells that spike... including the following!
pumpkin cells (from a stem)
skin cells
during egg fertilizaiton...</p>
<p>
rat pancreas,
paramecium...</p>
<p>
HH (Hodgkin-Huxley) formalism based on conductances...</p>
<p>
neurons r dynamical systems...</p>
<p>
can explain resonances, inhibition..</p>
<p>
his book dynamical systems in neurosci...</p>
<p>
he shows how good his model is... (cannot tell model from real data)</p>
<p>
we know HOW spikes are generated. but not WHAT they r for!!!</p>
<p>
spiking is reduced to firing rate by most neurosci... Big Mistake!</p>
<p>
firing rate... most dangerous misconcep in neurosci...</p>
<p>
muscles may not care abt spike timing (probably)</p>
<p>
coding of tactile stimuli...</p>
<p>
must saccade or touch to know where things are...</p>
<p>
he shows LGN responses in cat
2 cats years apart.. shown a movie... spiking timing is exactly the same!
reinagel and clay reid 2002</p>
<p>
simon thorpe.. RSVP rapid series visual processing...</p>
<p>
STDP obviously depends on spike timing</p>
<p>
combinatorial richness of encoding...
can downstream neuron tell the difference...</p>
<p>
axonal conduction delays...
in rabbit its perhaps 50 ms</p>
<p>
1/4 of neurons are typically doing nothing..
avg firing rate = 0.1 hz...</p>
<p>
if you insert one extra spike into his whole brain model in his PNAS paper then the behavior of the model
diverges in a quarter of a second (inserting only one extra spike!!) .</p>
<p>
every spike matters</p>
<p>
but peter lathma answered in a letter</p>
<p>
sensit to perturbations in vivo implies high noise and suggests rate coding in cortex!!</p>
<p>
nature july 2012</p>
<p>
he cites a study in which one single spike in a nerve => feeling in finger...</p>
<p>
genl anesthetics... scramble timing... perhaps</p>
<p>
when scramble conduc delays... the groups all disappear...</p>
<p>
vesicle release is probabilistic....</p>
<p>
sureya... do u compute in a chaotic system with lots of noise?</p>
<p>
polychronous system... precision is embed cuz of strong connectivity...
</p>