@article {pub3022pub3094,
	title = {Optimal Local Rejection for Classifiers},
	author = {Lydia Fischer AND Barbara Hammer AND Heiko Wersing},
	year = {2016},
	abstract = {We analyse optimal reject strategies for classifiers with input space partitioning, e.g. prototype-based classifiers, support vector machines or decision trees based on real-valued, distance-based certainty measures such as the distance to the closest decision border. We compare reject schemes with global thresholds, and local thresholds for the partitions of the space induced by the classifiers.
For the latter, we develop a polynomial-time algorithm to compute optimal thresholds based on a dynamic programming scheme, and we propose an intuitive linear time, memory efficient approximation thereof with competitive accuracy. Evaluating the performance in various benchmarks, we conclude that local rejection is beneficial in particular for simple classifiers, while the improvement is less pronounced for advanced models. An evaluation for biomedical data highlights the benefit of local thresholds.},
	publisher = {Neurocomputing},
	journal = {Neurocomputing},
	volume = {214},
	pages = {445-457}
}