[Mark]

This quote by David Olmsted is from the “Application to computer game playing (Chess/Go/...)� thread. The topic was starting to get away from game playing so I’d like to give it a fresh start if I may:

[DavidOlmsted]

The brain does not have any clock signal like computers that tell when the state of the brain is supposed to transition into another state. Yet state transitions can be thought of as discrete and as occuring instantaneously. Mathematically, this is represented by impulse functions. Amazing how impulse functions are similar to action potentials .
_________________
Click to go to my site at neurocomputing.org

[andy.goryachev]

It appears that the brain may be "clocked", but not like a CPU is clocked with a single clock signal. There seems to be a number of various periodic rhytms (alpha wave for instance).

I wonder if HTM implementation exhibits the same oscillating activity patterns.

[chatham]

the last poster is correct - there are many oscillatory periodic rhythms in cortex, some of which are thought to be especially important for attention and memory. foremost among these are the gamma and theta oscillations, which are thought by some (Jensen, Lisman and Idiart) to interact in a kind of "multiplexing mechanism," by which a specific pattern of neural firing may preferentially occur during a particular phase of the theta cycle. It appears that there are around 5 preferntial firing slots within the phase of a single theta cycle, which corresponds roughly to what one would expect to happen if information was encoded by a multiplexed gamma/theta rhythm.

Check it out here:

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=16161035&dopt=Abstract
_________________
http://scienceblogs.com/developingintelligence

[joel]

I think the following model might apply, at least in some cases...

1)A network receives input while in a 'low-gain' mode.

2)The input path is blocked, and the 'gain' is increased, causing the network to make its decision.

repeat...

W. Freeman seems to have observed this cycle in rabbit olfactory bulb, paced at the rabbit's resperatory rate. I think others have claimed to observe this in the hypocampus/cortex hookups, running at EEG-type rates.

Anyways, EEG itself demonstrates that there is lots of coordinated sequenced activity in the brain, whatever that activity might be. But calling this a clock as in the context of computers would be misleading. At least I hope it's not the same thing, as getting every bit of a computer to sync up to a GHz clock and never miss a bit is a big PTA.

[DavidOlmsted]

Individual neurons can be simulated to pulse with a very regular pattern for timing control but this is not the same as saying the brain is clocked. Clocking defines when one system state transforms into another according to fiinite automata theory. Since this is the only information processing theory published how can the brain process its information without having discrete states?
_________________
Click to go to my site at neurocomputing.org

[joel]

>>Individual neurons can be simulated to pulse with a very regular pattern for timing control

You'll have to tell me your learning rule. I practice and practice with my metronome, but still can barely keep a steady beat! Cheers!

[Nugenta]

I would say don’t think about the "neural states" but rather think in terms of synaptic states. A neurons output pulse may take on a huge variety of patterns based on whatever inputs it receives. The questions is if the underlying synaptic states are changing. With a consistent environmental structure, a learned system shouldl converge on a stable synaptic state irrespective of it being a clocked system or asynchronous. And a neuron is not much more than a group of synapses.

~Alex