Introduction to Artificial Neural Network Seminar Topic:
There are few traits of humans such as hearing, learning, reading and face recognition that are not yet mastered by computers. This is because of complexity associated in evaluating the data by the computers to do the above mentioned tasks. For the last 50 years developers have taken steps to achieve it and this field is termed as Artificial Neural Network. The name is derived because to evaluate results programmers had used software which uses a network of processors which reads the data and compare with each other and finally give the result.
BERNARD WIDROW from Stanford University had done a great deal of work in the field of neural network since 1950’s.
Neural network systems help in formulating an algorithmic solution, assessing the behaviour of a subject, selecting a structure from the source/existing data.
In Artificial neural network, programs are designed to evaluate any problem by duplicating the structure as well as functions of human nervous system. Simulated neurons form the base of neural network. These neurons are joined together in a different of ways to create networks. There will be resemblance of human brain and neural network.
There are 7 major components of artificial neurons: –
- Weighing factors
- Summation Function
- Transfer Function
- Scaling and Limiting
- Output function (competition)
- Error function and back-propagated value
- Learning function – Supervised, unsupervised and Reinforced.
After neural networks are programmed to get the desired results the network has to be trained accordingly. The training is done in different ways such as altering the interconnection weights and on what criteria the weights have to be changed in response to the inputs.
When it comes to applications artificial neural networks are used in many fields like Aerospace, automobile control, banking, credit card activity checking, defence electronics, entertainment, finance, industrial, insurance, manufacturing, medical, oil & gas, robotics and many other fields. There are few disadvantages such as learning time and complex implementation.
Download Artificial Neural Network Seminar Report.