Scientific Understanding of Consciousness
Consciousness as an Emergent Property of Thalamocortical Activity

Neural Network — Stochastic Behavior

 

 

Almost Poisson nature of the spike trains of single neurons found in most brain areas. (Rolls; Memory, Attention, and Decision-Making, 72)

Spike trains are essentially Poisson-like because the cell potential hovers noisily close to the threshold for firing, the noise being generated in part by the Poisson-like firing of the other neurons in the network. (Rolls & Deco; Noisy Brain, 78)

Noise and spontaneous firing help to ensure that when a stimulus arrives, there are always some neurons  very close to threshold  that respond rapidly, and then communicate their firing to other neurons through the modified synaptic weights, so that an attractor process can take place very rapidly. (Rolls & Deco; Noisy Brain, 78)

Noise inherent in brain activity has a number of advantages by making the dynamics stochastic, which allows for many remarkable features of the brain, including creativity,    probabilistic decision making,    stochastic resonance, unpredictability, conflict resolution, symmetry breaking, allocation to discrete categories, and many of the important memory properties. (Rolls & Deco; Noisy Brain, 80)

Computations can be performed through stochastic dynamical effects, including the role of noise in enabling probabilistic jumping across the barrier in the energy landscape describing the flow of the dynamics in attractor networks. (Rolls & Deco; Noisy Brain, 81)

 

Stochastic nature of Neuronal behavior

Because single neurons have small and uncertain effects on other neurons, the cortical description must be carried out in terms of neuronal populations rather than at the level of individual cells. (Stevens; Cortical Theory, 242)

Layer 4 neurons have a dendritic tree with a diameter of about 0.3 mm. (Stevens; Cortical Theory, 242)

Layer 4 is about 0.3 mm thick, and cortex has a density of about 10⁵ neurons per mm³. (Stevens; Cortical Theory, 242)

All of the neurons in layer 4 that fall within a cylinder having a radius of about 0.3 mm will have overlapping dendritic trees. The number of neurons that overlap is estimated to be approximately 8000. A significant fraction of this population of neurons should represent essentially the same information. (Stevens; Cortical Theory, 242)

A given axon generally arborizes over a considerable region of cortex with an arbor diameter of perhaps 0.5 mm, and forms about 2000 boutons, each of which makes one or two synapses. (Stevens; Cortical Theory, 242)

Neurons of the same functional class and in the same cortical layer share nearly the same potential synaptic inputs whenever their cell bodies are separated by several hundred microns or less, and the degree of similarity in their inputs increases as the distance between cell bodies decreases. (Stevens; Cortical Theory, 243)

Neuronal ensembles, not individual neurons. (Merzenich; Neural Representations, 65)

The reliability of spike transmission increases steeply for approximately 20 to 40 synchronous thalamic inputs in a time window of 5 milliseconds, when the reliability per spike is most energetically efficient. The optimal range of synchronous inputs is influenced by the balance of background excitation and inhibition in the cortex, which can gate the flow of information into the cortex. Ensuring reliable transmission by spike synchrony in small populations of neurons may be a general principle of cortical function.

Research study — Spike Synchrony in Small Populations of Neurons

 

The overall stochastic nature of neuronal behavior suggests that the physiologically meaningful signal from cortex should be the average firing rates of a population of perhaps 100 to 1000 neurons near a particular cortical site. (Stevens; Cortical Theory, 243)

The behavior of cortex at a particular point would then be described by the firing in a population of neurons. The total firing that represents this population would be determined by a weighted average of the appropriate neurons in the cortical region that surrounded the point, perhaps with weights that are described by a spatial Gaussian. Moving from one cortical location to an adjacent one, the variables describing cortical state would vary continuously with cortical position. (Stevens; Cortical Theory, 243)

A theory of cortex must be coarse-grained and treat cortical inputs and outputs as continuous variables that represent the summed behavior of appropriately sized and selected neuron populations. (Stevens; Cortical Theory, 243)

The prominent recurrent nature of lateral intracortical connections and relatively wide spatial distribution of cortical inputs mean that the cortical output at any one location must depend on both the input and output over relatively great expanses of cortex. That is, the output at any one point must be a functional, of both inputs and outputs. (Stevens; Cortical Theory, 243)

Efficacy of transmission of cortical pathways is modulated by the activity of control neurons whose primary functionality is to dynamically route information through successive stages of the cortical hierarchy. (Van Essen; Dynamic Routing Strategies, 276)

 

 

Return to — Neural Network