Methods and logic for autonomous generation of ensemble classifiers, and systems incorporating ensemble classifiers

A method for generating an ensemble classifier, the method comprising:

• transforming, automatically with a processor, multidimensional training data into a plurality of response planes according to a plurality of recognition algorithms, wherein each of the response planes comprise a set of confidence scores;
• transforming, automatically with the processor, the response planes into a plurality of binary response planes, wherein each of the binary response planes comprise a set of binary scores corresponding to one of the confidence scores;
• transforming, automatically with the processor, a first combination of the binary response planes into a first set of diversity metrics according to a first diversity measure;
• transforming, automatically with the processor, a second combination of the binary response planes into a second set of diversity metrics according to a second diversity measure;
• selecting a first metric from the first set of diversity metrics;
• selecting a second metric from the second set of diversity metrics;
• generating, automatically with the processor, a predicted performance of a child combination of the recognition algorithms corresponding to the first combination and the second combination, wherein the predicted performance is based at least in part upon the first metric and the second metric;
• selecting parent recognition algorithms from the recognition algorithms based at least in part upon the predicted performance;
• and
• generating the ensemble classifier, wherein the ensemble classifier comprises the parent recognition algorithms.