@inproceedings {pub3641,
	title = {Mitigating Concept Drift via Rejection},
	author = {Barbara Hammer AND Heiko Wersing AND Jan Philip Goepfert},
	year = {2018},
	month = {October},
	abstract = {Learning in non-stationary environments is challenging, because in them the common assumption of independent and identically distributed data does not hold; when concept drift is present it necessitates continuous system updates. In recent years, several powerful approaches -- such as ensemble techniques or intelligent memory models -- have been proposed and are able to deal with different types of drift. However, these models typically classify any input, regardless of their confidence in the classification -- a strategy, which is not optimal, particularly in safety-critical environments where alternatives to a (possibly unclear) decision exist, such as additional tests or a short delay of the decision. Formally speaking, this alternative corresponds to classification with a reject option, a strategy which seems particularly promising in the context of concept drift, i.e.~the occurrence of situations where the current model is wrong. In this contribution, we investigate learning under concept drift with a reject option. Specifically, we introduce an extension of the recently proposed self-adjusting memory architecture on top of a nearest neighbor classifier (SAM-kNN) to classification with a reject option, and we evaluate its performance in comparison with state-of-the-art alternatives in a variety of benchmark and original problems.},
	publisher = {Springer},
	booktitle = {International Conference on Artitificial Neural Networks ICANN 2018}
}