The study of Artificial Neural Networks or Simulated Neural Network has gained rapid growth due to the raise in computation speed. Researchers felt a strong need to develop computers that can work even half as good as the amazingly complex human brain. Human brain has billions of cells known as neurons that are interconnected in the form of mesh network. These brain cells are used to carry/transfer information. Computer scientists felt the need of devising a computer system that can work as fast as a neuron. This study is framed as Artificial Neural Networks (ANN) because it caters artificial methodologies that intends to mimic the Natural Neural Network which is that of the human brain.
It’s being said that today’s silicon ICs (Integrated Circuits) have processing speed up to nano-seconds (10^-9 seconds) whereas a human brain can process in milli-seconds (10^-3 seconds), which means human brain is 5 to 6 order slower than the computation speed of that of the computer. A question that needs to be addressed here is then how is it that the neural processing is much faster than the computer. This is because the neural network of human brain is massively parallel i.e. comprising of 10 billion nerve cells/neurons and around 60 trillion interconnections.
Structure of a human brain nerve cell or neuron
Axons are the transmission lines/carriers, which contain high Resistance and high Capacitance. Synaptic terminals constitute the response/output zone. Dendrites are connected to the Synaptic inputs.
Structure of the artificial neuron
Comprise of a summation function, which generates the sum of all the signals send to it. W1 , W2, …. Wn are the strength of the signals which are altered in order to change the desired result on the output.
X1.W1 + X2.W2 + …. + Xn.Wn = summation on the neuron —- Eq.(A)
We don’t require a linear result in this regard i.e. Eq.(A) rather we require a quantifiable unit. So a non-linear processing unit is required just before the output. To obtain a more quantifiable response.