Reduction Techniques for Instance-Based Learning Algorithms

By D. Randall Wilson, Tony R. Martinez, Robert Craig Holte. 2000

1,298 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Using evolutionary algorithms as instance selection for data reduction in KDD: an experimental study (2003, methodology): “[33]—Similar to Rmhc, it randomly selects a subset from TR, which contains a fixed number of instances ( ) but in each iteration the algorithm interchanges all instances from with instances from TR.”
Class Conditional Nearest Neighbor and Large Margin Instance Selection (2008, methodology): “In instance selection, prototypes are constrained to be members of the training set, and the selection process is typically based on heuristic search (see for instance [42]).”
Enhancing the effectiveness and interpretability of decision tree and rule induction classifiers with evolutionary training set selection over imbalanced problems (2009, background): “In the specialized literature, we may find some papers for resampling techniques from the point of view of the study of the effect of the class distribution in classification [41,16] and adaptations of instance selection methods [44,33] to treat with imbalanced data sets.”
Evolutionary Undersampling for Classification with Imbalanced Datasets: Proposals and Taxonomy (2009, methodology): “286 Evolutionary Computation Volume 17, Number 3 • Decremental reduction optimization procedure 3 (DROP3; Wilson and Martinez, 2000) is based on the rule “Any instance incorrectly classified by its k-NN is removed.””
Adaptive Machine Learning for Credit Card Fraud Detection (2015, background): “Since noisy samples are likely to be misclassified, many of them will be added to the S set which means CNN rule is extremely sensitive to noise [99].”
Neighborhood classifiers (2008, neutral): “…to deal with these problems (Anil, 2006; Fu, Chan, & Cheung, 2000; Fukunaga & Narendra, 1975; Hart, 1968; Kuncheva & Lakhmi, 1999; Kushilevitz, Ostrovsky, & Rabani, 2000; Lindenbaum, Markovitch, & Rusakov, 2004; Short & Fukunaga, 1981; Vidal, 1986; Wilson & Martinez, 2000; Zhou, Yan, & Chen, 2006).”
A review of instance selection methods (2010, methodology): “The border instances of a class provide useful information for preserving the class discrimination regions (Wilson and Martínez 2000; Brighton and Mellish 2002); therefore some filter methods are focused on selecting border instances.”
Supervised machine learning: A brief primer (2020, background): “Label noise has been shown to lead to inaccurate predictor selection and rankings (Frénay et al., 2014; Gerlach & Stamey, 2007; Shanab et al., 2012) and to decrease prediction performance (Lachenbruch, 1966; Nettleton et al., 2010; Wilson & Martinez, 2000).”
Classification in the Presence of Label Noise: A Survey (2014, methodology): “Classiﬁcation performances of the k nearest neighbors ( k NN) classiﬁer are also affected by label noise [94], [95], in particular when k = 1 [96].”
Experimental study on prototype optimisation algorithms for prototype-based classification in vector spaces (2006, result): “Therefore, as a general rule, and according to the previously published results [10, 5 ], the Wilson’s editing is considered to properly remove a possible overlap between the classes.”
Machine learning: a review of classification and combining techniques (2006, background): “Brighton and Mellish (2002) found that their ICF algorithm and RT3 algorithm (Wilson and Martinez 2000) achieved the highest degree of instance set reduction as well as the retention of classification accuracy: they are close to achieving unintrusive storage reduction.”
Evolutionary stratified training set selection for extracting classification rules with trade off precision-interpretability (2007, methodology): “The instance selection algorithms select representative instance subsets following a determined strategy, and they can improve the nearest neighbour rule prediction capabilities used in some cases as selection strategy objective [6,7].”