Mohit Sewak

.

.

Neural Networks

.

Our brain consists of approximately 10¹¹ neurons, in a densely connected and functionally related in a peripheral nervous system. These neurons are biological cells which get energized by the reaction of chemical ions, which generates an electrical signal which then propagates the neural network. These neural networks are nothing but a network of all of these 10¹¹ neurons, in which each of these neuron is interconnected to several thousands of other neurons. The axon (output node) of one neuron is not connected to the dendrites (input node) of other neurons, through synapses (connectors between input and output nodes).

.

.

The excitation of the neurons by chemical ions causes electrical spikes that travels along the axon of the pre-synaptic neuron (transmitter neuron) that further triggers the release of neurotransmitter substances at the synapse. The neurotransmitters cause excitation or inhibition in the dendrite of the post-synaptic neuron (receiver neuron). The integration of the excitatory and inhibitory signals may produce spikes in the post-synaptic neuron. The contribution of the signals depends on the strength of the synaptic connection. Through a network of these electrical signals and flows, various senses and motor nerves are energized and and information flows from one part of the nervous system to another.

.

.

Although at a unit level the system is very simple, but the very complexity of the interconnections and the huge number of the network nodes makes possible the accomplishment of extraordinary complex tasks through these networks.

.

.

.

.

.

.

.

.

Power of Biological Neural Networks in Different Species

.

.

Numerous experiments, and researches have been conducted over how we remember, identify and recall things. It has often been seen that once we see something, we are reminded of something else. This concept is widely (although without our knowledge) often being used by advertising and publishing companies to associate their brands with something that has a distinctly positive connotation in our lives.

For example, Coca-Cola with its widely popular advertisement slogan “Thanda bole to Coca Cola”, has attempted successfully to link Coca Cola with the concept of “Cold Drinks” literally. So that the next time you think of having a cold drink, you can instantly recall Coca Cola. There has been numerous such examples, but the bottom line is that there are some networks/ associations in our brain that have the capability of linking two different concepts residing in two different (sometimes seemingly unrelated) corners of the brain.

This can be done either by enforcing an already existing connection (axon-synapse-dendrite) between two nodes (neurons), or (re)creating a connection which (has been broken)never existed earlier.

.

.

.

There have been several experiments with birds and animals as well which further proves this concept of the working of memory, identification, association and recall through the processes in our biological networks. One of these is:

.

.

The Pigeon in Skinner Box Experiment:

.

In this experiment pigeons are made to recognize the paintings of two different artists (here Chagall, and Van Gogh). First they are shown the same paintings that they were shown earlier (when the distinction was made between the artists and their paintings), and then they were shown an entirely different sets of paintings of these two painters which has never been shown to these pigeons earlier. the results were:

.

1. In the first case the pigeons could distinguish between the paintings of the two artists with 95% accuracy.

2. In the second case of previously un shown paintings the accuracy was still remarkable high with 85%.

.

.

Implications and Uses of the findings of the capabilities of the Neural Networks:

.

The accuracy of 85% (for unknown, uncertain cases) is quiet high than what is possibly with any computer algorithm present (being used). Also the working of these neural transmitters (at the unit level) is quiet simple, and as they work on impulses of electric signals, so is also replicable in today’s sophisticated electronic chips.

.

.

It is only the volume, and the complexity of these networks that make them solve such complex and typical problems that seems impossible for a single neuron, or even a human to solve. And computers (especially with their increasing processing power and memory) are ideally suited to be programmed for such voluminous, fast (yet simple) inter-connective transmissions to solve a problem similar to how a human brain solves them.

Thus we come to the concept of ARTIFICIAL NEURAL NETWORKS, which will be covered in greater detail in the subsequent articles.

.

.

.

.

.

.

.

.

.

.