Neural Networks in Nature

The human brain is one of the most complex structures known to mankind, being made up of billions of interconnected cells (known as neurons). The most important part of the brain from our point of view is the Cerebral Cortex which is responsible for controlling motor functions (movement of muscles etc) and processing sensory data (the most complex of which are sight, hearing and touch). The cerebral cortex also contains the groups of cells associated with thought processes.

The cerebral cortex is a thin “skin” around the outside of the brain which is only a few cells thick. The surface of the brain is creased and folded in order to increase its surface area and thus the size of the cerebral cortex. Figure 1 shows a schematic diagram of the brain with the areas associated with various important functions highlighted.

Lateral view of the human brain (facing left)
Figure 1: Lateral view of the human brain (facing left)

The basic building block of the brain is the neuron. Neurons are a special type of cell (similar to nerve cells) which are capable of processing electrical signals. Neurons are connected together by long fibres along which electrical signals travel. The fibres are linked to neurons via a small gap known as asynaptic connection or synapse. We can think of neurons as simple electronic components and synapses as the wires which link them together.

Figure 2 shows a diagram of a neuron. Synapses on the left hand side form the neuron’s many inputswhich link to the cell’s dendritic tree. Synapses on the right are connected to the neuron’s single output, the axon. The strength of electrical signals passing through a synapse is altered slightly just before reaching the neuron. Unlike electronic components, neurons can not be simply soldered together, so the incoming signals must jump accross a small gap: the synapse (shown as a black circle at the end of the incoming nerve fibres). Each different synaptic connection purturbs electrical signals differently.

Schematic of a neuron
Figure 2: Schematic of a neuron

The output of a neuron is transmitted to other neurons via the cell’s axon and synaptic connections. Neurons are said to fire when their output oscillates at around 100Hz. Neurons can be either firing or inactive and thus their
output is effectively boolean (on/off).

Synaptic links forming the inputs to a neuron can be either excitory or inhibitory. Excitory inputs encourage the neuron the fire whilst inhibitory inputs discourage firing. The neuron takes all synaptic inputs into account, firing only if the magnitude of its summed inputs is greater than a threshold value.

The main point of note is that the output of a neuron is the result of a simple operation based only on that neuron’s inputs.

The power of a neural network (be it biological or atrificial) comes from the way in which simple building blocks (neurons and synapses) are linked together to form a more complex system. Essentially a neural network is greater than the sum of its parts.

The cerebral cortex has a remarkably simple structure, which has provided the inspiration for the structure of its artificial counterparts. We will use the visual cortex as an example here.

Layering of Cerebral Cortex
Figure 3: Layering of Cerebral Cortex

The visual cortex performs processing on raw data which it receives from the eyes, via the optic nerve. The optic nerve connects directly to an area on the rear of the brain. As shown in figure 4, the connections from nerves are linked to neurons on the outer surface of the cerebral cortex and electrical signals are then filtered through a number of layers of cells before being sent off to other parts of the brain for further processing. The neurons in each layer perform simple operations on data they receive from the layer above – each one further processing and translating.

To summarise…

  • The brain is built entirely from neurons
  • Fibres link neurons together like wires
  • Fibres are joined to neuron inputs via synaptic connections (synapses)
  • Synapses alter the magnitude of electrical signals slightly
  • Neurons accept many inputs but give only one output
  • The output of a neuron is a simple function of its inputs
  • A neuron can be said to be either firing or inactive
  • Neurons are often found in layered structures

Leave a Reply

Your email address will not be published. Required fields are marked *