Neural Network: Giving a Definition
A neural network is a different technology which builds intelligent programs using models which is similar or parallel to the neurons of the human brain or it may follow methods that are found in genetic algorithm and artificial life. “Neural networks are appealing because they learn by example and are strongly supported by statical and optimization theories.” Neuron models of computation mimic the neurons which consist of a cell body having many protrusions which are branched called dendrites, and a branch called the axon. Here other neurons sent signals to the dendrites.
A neural network is a software (or hardware) simulation of a biological brain, which is also sometimes called an artificial neuron network or ANN. The whole process receives an input that computes some function of it, and the result thus computed will be passed along the network. The solutions are produced by parallel and distributed processing of the networks in neural connections, and also on threshold weights. NN has input, processing, and output layers. The user can only see the input and output layers, whereas the processing layer is not visible.
“A memory-based network that provides estimates of continuous variables and converges to the underlying (linear or nonlinear) regression surface is described. The general regression neural network (GRNN) is a one-pass learning algorithm with a highly parallel structure.”
In Artificial Intelligence, there are three machine learning techniques. They are Symbol Based, Connectionist, Social and Emergent.
The main theory behind Symbol Based is that symbols are used to reference objects and relations in a domain. In this connectionist is given more importance. In Connectionist or parallel distributed processing (PDP) is a neutrally inspired model, which deemphasizes the usage of symbols in problem-solving, where else Neural Networks takes the concepts as if systems of simple, interacting components have got intelligence, by the learning process or adaption by adjusting the connections between the components. Usually, Symbolic AI is a method of symbolically representing knowledge and inference, e.g. rules, semantic networks, frames, logic, etc. A connectionist approach is an alternative approach to it. But it is not may not be incompatible. It is a representation of an attempt to simulate the microstructure of the human brain, as the suggestion of neuroscientific researchers is that the brain is made of interconnected neurons.
“Two realizations of connectionist expert systems (shells) which facilitate building expert systems when raw data and/or expert rules are available are presented. The knowledge base is represented as a neural network trained either by using past data or using rules.”
The basis of neural networks is the artificial neuron. The cumulative input of the neuron is Σni=0wixi where w is the set of real-valued weights is the inputs.