Julian (and Laurent), Following David's suggestion, I replied on the nupic-theory list.
--Subutai On Sun, Jan 24, 2016 at 9:52 AM, Julian Samaroo <[email protected]> wrote: > Laurent, > > That paper and associated write-up are a bit exaggerated to say the least. > All it really states (of value) is that contrary to our belief that > synapses could only take on 2 or 3 distinct values (which has never really > been considered wholly factual), they can in fact take on many more values > (they "calculated" about 27-some possible states). However, IIRC, HTM > already uses continuous synaptic values (and if not, it probably should). > The "10X increase in memory" is really very hand-wavy, and is fully reliant > on the neuron's ability to discriminate between those now-discovered > additional synaptic states. The problem with that hypothesis is that > neurons already have very stochastic synapses, and the neurons themselves > are themselves quite noisy when integrating signals, so the ability to > utilize those minute differences in synaptic strength is very unlikely. Am > I saying it's not true, and the researchers are just bluffing? No, but I'm > saying that paper really doesn't have a very strong argument for its case. > > Julian Samaroo > > On Sun, Jan 24, 2016 at 1:48 AM, Laurent Julliard <[email protected]> > wrote: > >> Guys, >> >> I came across this article ( >> http://www.kurzweilai.net/memory-capacity-of-brain-is-10-times-more-than-previously-thought) >> and I was wondering if what they discover on synapse behavior could either >> improve in any way the current model of synapses in HTM and/or confirm the >> way synapses are potentiated today through the management of their >> permanence value ? >> >> -- >> Laurent Julliard >> Twitter @lrjay >> >> >
