Figure 2

Application of ANNs to problem solving: (a) pattern classification (i.e., assigning an unknown input pattern to any of prespecified classes based on properties that are characteristic to a given class); (b) clustering (i.e., clusters or classes are formed by exploring the similarities or dissimilarities between the input patterns based on their inter-correlations); (c) functional approximation or modeling (i.e., training an ANN on input-output data to approximate the underlying rule relating the inputs to outputs); (d) forecasting or predicting (i.e., training an ANN on samples for a time series [t(1) to t(n)] representing a certain phenomenon at a given scenario and then using it for other scenarios to predict the behavior at a subsequent time [t(n + 1)], and (e) association (i.e., developing a pattern by training an ANN to construct the corrupted or missing data).