Avoiding the interpolation inaccuracy in nearest feature line classifier by spectral feature analysis

1 In nearest feature line approach, the representational capacity of a given training set is 2 generalized by defining feature lines passing through each pair of samples belonging to the same 3 class. This technique is shown to provide superior performance on various classification problems 4 than the nearest neighbor approach. From the performance point of view, the major weakness 5 of this technique is the interpolation inaccuracy which occurs when a feature line passes through 6 samples that are far away from each other. Several variants are recently proposed to avoid this 7 weakness. In this study, we follow a different path and propose to transform the training data 8 of different classes into separate clusters before applying nearest feature line classifier. Spectral 9 clustering based transformation is used for this purpose and it is shown that the accuracies 10 achieved by both the nearest feature line and the shortest feature line segment approach which 11 is the most recent variant of the nearest feature line technique are improved. 12 ∗Corresponding author: E-mail: [email protected], Tel: +90 392 6302842, Fax: +90 392 3650711 1 keywords: Nearest feature line; Interpolation inaccuracy; Spectral clustering; Spectral feature 13 analysis; Shortest feature line segment 14