The emergence of polychronization and feature binding in a spiking neural network model of the primate ventral visual system.

Abstract

We present a hierarchical neural network model, in which subpopulations of neurons develop fixed and regularly repeating temporal chains of spikes (polychronization), which respond specifically to randomized Poisson spike trains representing the input training images. The performance is improved by including top-down and lateral synaptic connections, as well as introducing multiple synaptic contacts between each pair of pre- and postsynaptic neurons, with different synaptic contacts having different axonal delays. Spike-timing-dependent plasticity thus allows the model to select the most effective axonal transmission delay between neurons. Furthermore, neurons representing the binding relationship between low-level and high-level visual features emerge through visually guided learning. This begins to provide a way forward to solving the classic feature binding problem in visual neuroscience and leads to a new hypothesis concerning how information about visual features at every spatial scale may be projected upward through successive neuronal layers. We name this hypothetical upward projection of information the "holographic principle." (PsycINFO Database Record

A speed-accurate self-sustaining head direction cell path integration model without recurrent excitation

“…However, real neurons in the brain communicate with each other using discrete electrical pulses called action potentials or 'spikes'. In future work, we plan to run simulations of the head direction system using a spiking neural network model that can more accurately mimic the temporal dynamics of neurons and synapses (Eguchi et al 2018;Isbister et al 2018). In particular, this will enable us to implement realistic neuronal and synaptic time constants governing, for example, the exponential decays of both cell voltages and synaptic conductances.…”

Spike: A GPU Optimised Spiking Neural Network Simulator

“…A simulator whose performance is unaffected by axonal delays is particularly desirable given the range of research avenues which leverage delays within spiking neural networks. SNN models investigating a range of subjects including sound localization, reservoir computing, auditory processing and visual feature binding (Goodman and Brette, 2010;Paugam-Moisy et al, 2008;Erfanian Saeedi et al, 2016;Eguchi et al, 2018) are but a few of the many studies which implement multiple axonal delay values within a single study for modelling purposes. Spike is a simulator designed to optimise this process.…”

A new approach to solving the feature-binding problem in primate vision

“…How might the visual cortex represent such a hierarchy of visual features, as well as the hierarchical binding relations between these features, at every spatial scale and across the entire visual field? Eguchi et al [ 9 ] have recently shown how this may be achieved within a biologically realistic hierarchical neural network model of the primate ventral visual system with the following properties. (1) The model is a ‘spiking’ neural network, in which the timings of the spikes emitted by neurons are explicitly represented.…”

“… (5) The network may incorporate multiple synaptic connections between each pair of pre- and postsynaptic neurons, where these connections have different axonal transmission delays. Eguchi et al [ 9 ] showed that this allows the STDP to selectively strengthen specific synaptic connections with particular axonal transmission delays. …”

“…Using a neural network model with the above architectural components, Eguchi et al [ 9 ] reported that training the model on visual stimuli led to the emergence of repeating spatio-temporal patterns of spikes in the higher layers of the network. A subpopulation of such neurons that emit their spikes in a regularly repeating spatio-temporal chain is referred to as a polychronous neuronal group (PNG).…”