ABSTRACT
In many circumstances, the output of a signal-processing or image-processing algorithm can be sent directly to a human operator through some sort of graphics or image display. The problems associated with machine capacity and overtraining are examined in the context of the specific classification techniques described. Nonetheless, transforming the data into higher dimensions can still result in overfitting or overtraining, leading to poor generalization as well as extensive computational time. Classification searches for patterns in data sets, sometimes very large multivariable data sets, with the goal of associating these patterns with a limited set of classes. Linear classifiers attempt to construct boundaries that consist of straight lines, planes, or hyperplanes depending on the dimension of the data sets, that is, the number of measurement variables being considered. In many systems and classifiers, it is possible to trade-off between sensitivity and specificity.
