Untitled

Abstract

At the transition from early reptilian ancestors to primordial mammals, the areas of sensory cortex that process topographic modalities acquire the laminar structure of isocortex. A prominent step in lamination is granulation, whereby the formerly unique principal layer of pyramidal cells is split by the insertion of a new layer of excitatory, but intrinsic, granule cells, layer IV. I consider the hypothesis that granulation, and the differentiation between supra- and infra-granular pyramidal layers, may be advantageous to support fine topography in their sensory maps. Fine topography implies a generic distinction between "where" information, explicitly mapped on the cortical sheet, and "what" information, represented in a distributed fashion as a distinct firing pattern across neurons. These patterns can be stored on recurrent collaterals in the cortex, and such memory can help substantially in the analysis of current sensory input. The simulation of a simplified network model demonstrates that a non-laminated patch of cortex must compromise between transmitting "where" information or retrieving "what" information. The simulation of a modified model including differentiation of a granular layer shows a modest but significant quantitative advantage, expressed as a less severe trade-off between "what" and "where". The further connectivity differentiation between infra-granular and supra-granular pyramidal layers is shown to match the mix of "what" and "where" information optimal for their respective target structures.

Grid alignment in entorhinal cortex

“…This hypothesis, not developed in the present work, represents an interesting direction of modeling research, which could focus on how conjunctive cells and grid cells may develop together in different layers, and perhaps even self-organize into laminar structures 23,42 . From the point of view of our model, the best way to address the question of a single unit versus a population origin of grid alignment is to interfere with the head direction source of information, assuming with some plausibility that this signal is not directly generated in the entorhinal cortex.…”

“…The alignment of grid cells in layer II is perhaps inherited from the deeper layers, which send strong projections to superficial layers 41 . This hypothesis, not developed in the present work, represents an interesting direction of modeling research, which could focus on how conjunctive cells and grid cells may develop together in different layers, and perhaps even self-organize into laminar structures 23,42 .…”

Anti-correlations in the degree distribution increase stimulus detection performance in noisy spiking neural networks

“…However, the cortex is organized as a layered structure, generally thought to be comprised of functional cortical columns (Mountcastle 1998; Douglas and Martin 2004), which can improve computational efficiency beyond the capabilities of recurrent networks without such spatial organization (Treves 2003; Raizada and Grossberg 2003; Haeusler and Maass 2007). As we showed here, correlations in the degree distribution can also provide additional capabilities for stimulus detection.…”

Biologically-inspired neuronal adaptation improves learning in neural networks

“…Fontaine et al [ 26 ] studied membrane potential variability in barn owls, concluding that it is a genuine feature of neurons as opposed to a result of noise. Carandini et al [ 27 ] used a rectification model [ 28 ] to model the effect of adaptation in the cat visual cortex, and Treves [ 29 ] has discussed adaptation as a prerequisite for several key computational mechanisms throughout the brain [ 30–32 ]. Given the vast research on adaptation and its importance in the brain, here we suppose that adaptation may offer advantages for artificial neural networks and test its effect on networks trained to perform benchmark classification tasks.…”