Interlink Concept's Neural Network LOGO Story

Our logo is fashioned and represents a computer neural network concept.

A neural network is a computer program that models the way nerve cells (neurons) are connected together in the human brain. Neural networks enable a computer to train itself to recognize patterns in a strikingly human-like way. Like the human brain, neural networks give only approximate results, but they can do things that no other kind of computer program can do.

Each neuron has several inputs but only one output. Some of the inputs excite (activate) the neuron while others inhibit it, each with a particular strength. The idea is that each output neuron will be activated when one particular link of a pattern is present at the input. In the computer, the neurons and connections are simulated by arrays of numbers.

The Input Pattern of data enters on left and the Output Pattern exits on the right with the connections (each with its own weight) in between.

Training a neural network is like training an animal. Patterns are applied to the input and a simple algorithm adjusts the weights of the connections to try to get the desired output. After many training runs, in which this is done with many different patterns, the neural network "learns" to recognize patterns of a certain kind. Even the programmer need not know exactly what these patterns have in common. The patterns are analyzed by the neural network, not by the programmer.

Neural networks are good at recognizing inputs that are vague, ill-defined, or likely to contain scattered variation. For example, a neural network can recognize images of human faces, or patterns of weather data, or trends in stock market behavior. However, a neural network is never 100 percent reliable, and even simple calculations can be quit slow.

Source: Barron's Business Guides, Dictionary of Computer Terms, 3rd Edition, Michael Covington, Ph.D., Douglas Downing, Ph.D.