on the sixth day of creation... ורבו פרו and replenish the earth and subdue it, and have...
TRANSCRIPT
On the sixth day of Creation...
פרו ורבוAnd replenish the earth and subdue it,
and have dominion over the fish of the seaand over birds in the sky and reptilians on land
after some clarification...
granulate and multiply And replenish the…
GENESIS I, 28
granulate AND MULTIPLY ?!?
We took it to mean, to insert granule cell layers into our cortex
and then, to see if itleads to multiplication...
GRANULATE, I ….
take the medial wall of thepremammalian cortex
and insert the fascia dentata, with
its granule cells, at the input end
note that the granule cells havebecome (excitatory) interneurons
the new structure remains stable and unique across mammalian species
Multiply, I ?
NO!H
H
opossumhuman
GRANULATE, II
We took it seriously, and went on, trying to insert granule cell layers
into our cortex. Not quite in the
same way as for the medial wall...
• now, the dorsal cortex. It acquires fine topography ...
• ...and it laminates
granulating the dorsal wall, leads to the mammalian
isocortex
the brand new `neocortex’ has laminated, i.e. inserted a granular layer IV in between two pyramidal cells layers.
Layer IV granules are now (excitatory) interneurons
what does this other granulatio
n buy us?
Isocortical lamination
• emerges together with fine topographic mapping
• does not apply to the non topographic olfactory system
• is underdeveloped in caetaceans
It might be a computational solution to the
need to relay precise information about
both ‘where’ and ‘what’ sensory stimuli are.
sff
src
R
patch ofcortex
inputstation
input activity
spatial focus
detailed pattern
feedforwardconnections
recurrent collaterals
the model
The activation of units in the previous station is the product of a spatial
‘focus’, say, a Gaussian of radius R (which presumably would be picked up by
optical imaging, or by multi-unit recording) and a detailed unit-by-unit pattern of activity (which would require single unit recording to be revealed). p patterns of activity (e.g. 2-12) are established at the beginning, drawn at random from a given distribution, and used repeatedly in one simulation.
The activation of units in the cortical patch is compared with the activations resulting from the application of each input pattern at each spatial focus, to decode the pattern and focus x of the current activation. This allows measuring
as well as
both population measures, reflecting activity in the whole patch
)()(
),(log),( 2
decodedreal
decodedrealdecodedreal xpxp
xxpxxpposI
I
)()(
),(log),( 2
decodedreal
decodedrealdecodedreal pp
pp
identI
Both recurrent and feedforward weights are modified according to a
simple ‘Hebbian’ associative rule, over the course of several training epochs. Each training epoch involves presenting, in random order, each input pattern at each activation focus. The map is thus pre-wired at a coarse, statistical level, and self-organized at a finer scale.
After a training epoch, noisy versions, again of each pattern at each activation focus, are presented for testing, with no weight change. The full information about position and identity cannot be decoded from the activation in the patch, because the activation in the input is noisy (in practice, e.g. 40% of the input units follow the prescribed pattern, and 60% are randomly activated with the same distribution)
If R << Src, it is rather intuitive to predict how much information can be relayed by feedforward projections of spread Sff:
ffS
identI
ffSposI
)/1log(
• Iident is small initially
• grows with learning
• no difference between layers
Results for p=4
• Ipos is less affected
by learning
• decreases with more diffuse feedforward connections
• again, no difference between layers
These data, plotted as Ipos vs. Iident,
demonstrate the what/where conflict as a boundary
• using more patterns merely shifts the same boundary upwards
Differentiating a granular layer (IV)
in which units receive focused FF connections, also more restricted RC connections, and follow a specific dynamics
• may nail down the focus of activation within the cortical map (preserving detailed positional information)
• without interfering with the retrieval of the identity of the specific activation pattern (achieved mainly by the
collaterals of the pyramidal layers)
sff
src
R
patch ofcortex
inputstation
input activity
spatial focus
detailed pattern
feedforwardconnections
recurrent collaterals
the model
Indeed it happens!
Laminated cortex canrelay more combinedwhat and whereinformation than if it
were not laminated
• The advantage is somewhat more evident for larger p
• it is small, but should scale up in a network of realistic size
Dependence on the size of the cue: the effect of learning...
…the advantage is there whatever the size of the cue
2) focus its afferents
but what do I do to layer IV ?
1) restrict its collaterals
3) sustain its dynamics(but suppress it in training)
The granular layer
may nail down the focus of activation within the cortical map (preserving detailed positional information)
without interfering with attractor-mediated retrieval of the identity of the specific activation pattern (achieved mainly by the collaterals of the pyramidal layers)
A differentiation between supra- and infra-granular layers may be usefully coupled to their different extrinsic connectivity, if:
• the supragranular layers preserve both positional and identity information, and trasmit it onward for further analysis
• the infragranular layers relay backwards and downwards identity information freshly squeezed from the attractors, without bothering to replicate positional information
and what do I do to layer V ?
4) remove its afferents from layer IV
V
III
IV
Lamination+direction
al connectivity make each layer convey a better mix of information, beyondthe capability of any unlaminated patch,whatever its Sff
• they also slow down learning, though, so the advantage would be greater if more learning epochs had been allowed (here they are set to 3)
Oops! I forgot the timing..
..this account is roughly independent of dynamics (a detailed analysis of relative timings, e.g. of the different inputs to the deep layers)
the only “dynamical” element introduced is firing frequency adaptation, which is however used in a time-independent fashion
we shall discuss more time-related uses of adaptation over the next two days, in generating transitions along continuous and among discrete attractors.
A functional hypothesis
A common mode of operation of the primordial sensory neocortex of mammals may have been autoassociative attractor dynamics.
Attractors may be formed by self-organizing weight changes on FF and RC connections, and may dominate the dynamics of both SG and IG layers, although the former can be kept in tighter positional register by layer IV.
Thanks to Hamish Meffin, with whom I discussed such ideas, with divergent conclusions (see his Ph.D. Thesis, U. of Sidney)
2 suggestions
• Understanding specific mammalian mechanisms of information representation and retrieval may require quantitative (information theoretical) analyses at the level of populations of individual neurones
• Only notions of sufficient abstraction and generality as to apply to each sensory cortex can help explain the appearance, in evolution, of this universal neocortical microchip.
Multiply, II ?
We are busy trying to understand it. Maybe next time...
but why ?
YES !
hedgehog
cat
monkey