Using dynamic expansion and contraction approach to generate a feedforward network

This paper proposes an algorithm to autonomously generate a feedforward network based on the input data. Dynamic expansion and contraction approach (DECA) is used to determine the optimal number of hidden nodes. The algorithm is applicable to the neural network used for function approximation and pattern classification. The short interval of train/test interleaving will minimize the learning time and avoid over-training the network. That is, a neural network for an application can be generated automatically in optimal time. Together with time-division-multiplexing architecture a hardware reconfigurable ANN architecture with learning capabilities can be realised