Abstract:
During the last decade, ANNs have emerged as a powerful tool for pattern recognition and modeling output from a system using the input data. Artificial Neural Networks are relatively crude electronic models based on the neural structure of the brain. The application of ANNs to water resources problems has become popular due to their immense power and potential in modeling non-linear systems. Among the many ANN structures that have been studied, the widely used structure in the area of hydrology is the multi-layer, feed-forward network. Neural networks learn, they are not programmed. Yet, even though they are not traditionally programmed, the designing of neural networks does require a skill. This skill involves the understanding of the network topologies, current hardware, current software tools, the application to be solved, and a strategy to acquire the necessary data to train the network. This skill further involves the selection of learning rules, transfer functions, summation functions, and how to connect the neurons within the network.
