What is Neural Networks?
Introduction to Brain
- Made up of very large number of simple processing elements
- 10^11 neurons and 10^15 dynamic interconnections among them..
- Excellent fault tolerance.
- Highly parallel processing
- Capacity to deal with vague environment.
- Very good at pattern matching.
- Good at solving ill defined problems particularly perception related ones
- Highly parallel and distributed control
- Content Addressability
- No loss of information while neurons are constantly dying
- It learns through examples
- Only important information is kept on-line
- Being perfect by practice
- Can apply previous Knowledge to solve newer but slightly different problems
Neural networks, what are they and why they are to be studied
- Realization of mathematical model of brain in programming
- Different models hence different NNs
- They are applicable to problems earlier not possible to be solved by computer.
- They are to brain what airplanes are to birds
Applications of Neural Networks
- Simulations and Neuro-boards
- Vehicular control
- Battlefield management
- Speech generation and to some extent understanding.
- Undersea mine detection
- Financial Analysis
- Airport Explosive detection
History of Neural Networks
- Birth in mid 50.(Perceptron)
- EXOR Problem made Known by Minskey and Papert.
- Eclipse of almost two decades
- Work of Turvo Kohonen, Stephen Grossberg etc.
- Era of multilayer Perceptrons and success for Neural Networks.
Natural Neuron