Artificial Consciousness/Neural Correlates/Neural Models/Dendrite Model

The Dendrite Model
One way of looking at a Neuron is as a data processing device. Data comes in one end, gets processed somehow in the middle and goes out the other end. While this might be a little simplistic if only because it assumes that there are not outputs in the input end of the device, or inputs on the output end of the device, both of which have been detected in neurons, it is safe to say that by and large, Neurons act in a manner similar to this model, so it has some value. Of interest to this Model, is the Input end of the Data Processing Device.

Theoretically, a number of devices can impinge on the input of a data processing device. In reality, the number may be limited by the technology used by the device. For instance in electronics it is not surprising to find that only so many devices can attach to any one input before the electrical characteristics of that input destroy the utility of the device.

Theoretically again, there is no difference between inputs to a device, they can all be modeled as being the same, in reality however, it is quite often the case that different devices require different interfaces, so not every device can connect to every other device. However theoretically we could symbolize an unknown network of connections as being a star that centers on the input of the device, and consider the differences in interface simply to be details of the connections needed to connect the devices. We call this abstraction an Instar because it is on the input side of the device.

Another way of looking at it, is that we might need a network of connections to gather signals from a number of devices, and so the Instar is a gathering network. In reality this is perhaps the better model, in that real neurons do not have infinite branching to allow large numbers of inputs to be integrated at par, and so some connections must travel longer distances to attach to the device, resulting in detectable delays in signal transfer. This is especially true because the electrical characteristics of neural processes are variable and depend on factors such as the size of the process, the capacitance of the cellular membrane, and the number of ion channels that can act to amplify the signal along the length of the process. What is known is that far from reaching the speed of light like electrical impulses through wires, electrical impulses through neural processes are slow, and must overcome significant resistance and reluctance, with the result that they need significant amplification the further they must travel.

It seems reasonable therefore that the fastest reactions are going to be between signals that travel the shortest distances, and that therefore there is likely to be a tendency for some processing to happen within the network instead of waiting for transfer to the main device. This is probably why we have detected some pre-synaptic buds within the dendritic mass of neurons, they are there simply to allow processing within the dendrite mass instead of waiting for the signals to actually propagate long enough to reach the body of the neuron.

I will discuss how this changes the basic Hebbian Model later, but for now it is enough to know that some dendrites have pre-synaptic buds as well as sensitive patches, but their role is to act as inputs to the main device or neuron. One way of looking at dendrites is as incremental processing elements associated with delay lines. The incremental processing units or Syanpses, create a signal population that travels to the soma or body of the neuron where it combines with similar signals from other dendrites to form an integration or aggregate signal. Each time a dendrite branches we can see it as forming another sub-integration of the Neurons main integration, thus the often fractal shape of the dendrite forms a calculation. The delay line aspect of the model however means that signals that are gathered from different inputs, are combined not only spatially but temporally based on the distance they have to travel, and how large the process they travel through is, the result is that depending on the structure of the dendrite, a sensitivity to spatial structure or temporal structure to a signal might develop. An example where this might be important is if the structure of the dendrite resonates at a specific frequency, in which case any signals that have that frequency, would be amplified by their relationship making the neuron exquisitely sensitive to the frequency of the signal. Another example might be a range of neurons that are each tuned to a different structural frequency, that allows the sensitivity of touch that separates texture to be possible.