In our previous previous applications of ABML, we called this the ABML refinement loop .
In one of the first applications of ABML, the goal was to distinguish between a good and a bad bishop in a chess position .
We used ABML to learn a diagnostic model for distinguishing between different types of tremor in patients with a neurological disease , The patients were classified as essential tremor or parkinsonian tremor or mixed tremor (having both).
After arguments are added to the critical example, ABML relearns the model.
In ABML, however, too many reasons will result in poor generalization.
Although we encountered too specific arguments in almost every application of ABML, we have not yet used pruning.