Second‐Order Meta‐Algorithmics and their Applications
Steven J. Simske · 2013
This chapter shows the confusion matrix in action, and discovers how its confusion can indeed be spread to determination of precision, decision tree classification optimization, targeted classification, and ground truthing optimization. It provides more in-depth applications of the second-order meta-algorithmic patterns. All nine of the second-order patterns are demonstrated in the chapter, applied to the usual four primary domains: (1) document understanding, (2) image understanding, (3) biometrics, and (4) security printing. In addition, second-order patterns are applied to the domains of image segmentation and speech recognition. Second-order meta-algorithmics are in general more complicated than the first-order meta-algorithmics. Most of the patterns employ two or more decisioning approaches, enabling more design flexibility through trade-off of the parameterization of these multiple components. These patterns use a new set of analysis tools: confusion matrix, output space transformation, expert decisioners, and thresholded confidence. Controlled Vocabulary Terms image segmentation; speech recognition