Multi-Backpropagation Network

Neural Network is a computational paradigm that comprises several disciplines such as mathematics, statistic, biology and philosophy. Neural Network has been implemented in many applications; in software and even hardware. In most cases, Neural Network considered large amount of data, as it will be teach to learn or memorize the data as the knowledge. The learning mechanism for Neural Network is its learning algorithm. Backpropagation (or backprop) algorithm is one of the well-known algorithms in neural networks. Backpropagation network with hidden layer able to process and model more complex problem. However, as some problem involve a large amount of data, the network would be more difficult to train. More input units or hidden units could increase the model size and increase its computational complexity. Synonym to human learning, a complex problem required some time to learn or memorize. Therefore, reducing the network complexity would be an advantage to the network. This paper proposed multi-backpropagtion network to reduce the size of a large backpropagation network. The domain for the illustration presented in this paper is the Myocardial Infarction disease. This approach do not required any alteration of the algorithm. The large network is split into several smaller networks, which act as a specialized network. This approach could also reduce the redundant data and reduce the training epochs.