Back to the abstracts page
Back to the NCCD 2015 home page
Controlling the flow and transformation of information through neural circuits
Peter Stratton1,2, Francois Windels1 and Peter Silburn2
1The University of Queensland, Queensland Brain Institute (QBI), Australia 2The University of Queensland, Centre for Clinical Research (UQCCR), Australia

A basic understanding of how the brain computes currently eludes us. Experimental and theoretical studies suggest that certain characteristics of neural activity play essential roles in brain function. These characteristics are:

1. spike timing and propagation delays
2. oscillations and correlated spiking activity
3. self-organised criticality and balanced excitation and inhibition, and
4. complex, metastable dynamics.

We demonstrate that combining all of these principles into a single neural network model endows the network with the ability to fully control the flow and transformation of information -- that is, to perform computation.

We suggest that there are 7 basic computational operations (primitives) that the brain must implement in order to perform computation. They are: encoding, decoding, transformation, dynamic routing, long-term memory, short-term memory and binding.

To implement these functions, we introduce the concept of competitive cross-coupling (CXC). Briefly, the premises behind CXC are that information is actively represented in the brain by spatiotemporal patterns of spikes, and that these patterns compete globally across the brain through the diffuse thalamo-cortical matrix loop [1-3]. We devise a computational neural network model which uses CXC together with the above characteristics of neural activity (including metastable dynamics and non-uniform propagation delays between neurons) to demonstrate all of the necessary computational primitives [4]. Using this model, it becomes apparent that different spatiotemporal spike patterns can activate downstream neurons in complex combinatorial configurations. These complex configurations allow neural circuits to reliably but flexibly manipulate information, endowing the brain with its complex information processing capabilities.

The CXC model gives a mechanistic, neuron-to-network level account of the control of information in neural circuits, and may provide a framework for understanding how brain dynamics facilitate computation.

[1] Stratton, P. and J. Wiles, Self-sustained non-periodic activity in networks of spiking neurons: the contribution of local and long-range connections and dynamic synapses. NeuroImage. 52(3): p. 1070-1079 (2010).
[2] Stratton, P. and J. Wiles, Complex spiking models: a role for diffuse thalamic projections in complex cortical activity. Neural Information Processing. Theory and Algorithms: p. 41-48 (2010).
[3] Stratton, P. and J. Wiles, Global segregation of cortical activity and metastable dynamics. Frontiers in Systems Neuroscience (submitted), 2015.
[4] Stratton, P., F. Windels and P. Silburn, Computing with metastability: competitive cross-coupling (CXC) in neural circuits. Frontiers in Systems Neuroscience (submitted), 2015.