File Entry: Particle Swarm Optimization for creating self organized Evolutionary Back Propagation Artificial Neural Networks

A novel technique for developing a self  organized  Back  Propagation  Artificial Neural  Networks  (BPANNs)  or  simply  Back Propagation  Neural Network  (BPN) with  the 
help  of  swarm Intelligence  (SI) technique. This  paper  proposes  an  overlapping  swarm intelligence  algorithm  for  training of neural networks  in  which  a  particle  swarm  is assigned  to  each  neuron  to  search  for  that neuron’s  weights. The  training  of  ANN  is  a difficult task  when  it’s  get  into the  local minima and causes the low learning rates; it is required of an evolutionary multi dimensional algorithm  to seek  optimal weight  values, the positional  optimum  values  and  the dimensional  optimum  in  values  in  the dynamic  problem  space.  The  approach discussed throughout  in  this  paper  is  the credit assignment process by first focusing on updating  weights  and  biases  swarms Intelligence and then evaluating the fitness of the particles using a network. With the proper adaptation  of  the  SI-BPN  process,  the proposed  method  can  develop  an  optimum network  within  an  architecture  space  for  a particular problem. Additionally, it provides a class list of all other potential configurations. This algorithm will provide superior learning ability  to  the  traditional  Back-Propagation (BP) method in terms of accuracy and speed.