Artificial Neural Networks (ANN) are inherently parallel architectures which can be implemented in software and hardware. One important implementation issue is the size of the neural network and its weight adaptation. This makes the hardware implementation complex and software learning slower. In practice Back propagation Neural Network is used for weight learning and evolutionary algorithm for network optimization. In this paper a modified genetic algorithm with more fondness to mutation is introduced to dynamically evolve network structure and weights at the same time. A single layered feed forward neural network is designed and trained using conventional method initially, then the proposed mutation based modified genetic algorithm is applied to evolve the weight matrix and structure pruning of the neural network. This algorithm facilitates the hardware implementation of ANN.
cited By (since 1996)1; Conference of org.apache.xalan.xsltc.dom.DOMAdapter@3e94fe9c ; Conference Date: org.apache.xalan.xsltc.dom.DOMAdapter@27d7d3dd Through org.apache.xalan.xsltc.dom.DOMAdapter@4a96669a
Na Mohankumar, Bhuvan, Bb, NirmalaDevi, Mc, and Arumuga, Sd, “A modified genetic algorithm for evolution of neural network in designing an evolutionary neuro-hardware”, in Proceedings of the 2008 International Conference on Genetic and Evolutionary Methods, GEM 2008, Las Vegas, NV, 2008, pp. 108-111.