Adaptive, Hubness-Aware Nearest Neighbour Classifier with Application to Hyperspectral Data
We present an extension of the Nearest Neighbour classifier that can adapt to sample imbalances in local regions of the dataset. Our approach uses the hubness statistic as a measure of a relation between new samples and the existing training set. This allows to estimate the upper limit of neighbours that vote for the label of the new instance. This estimation improves the classifier performance in situations where some classes are locally under-represented. The main focus of our method is to solve the problem of local undersampling that exists in hyperspectral data classification. Using several well-known Machine Learning and hyperspectral datasets, we show that our approach outperforms standard and distance-weighted kNN, especially for high values of k.
KeywordsNearest neighbour Hyperspectral classification Hubness
This work has been supported by the project ‘Representation of dynamic 3D scenes using the Atomic Shapes Network model’ financed by National Science Centre, decision DEC-2011/03/D/ST6/03753. Authors would like to thank Marcin Blachnik for extended discussion on the first version of the paper and Krisztian Buza for his insightful comments and for making available the PyHubs (http://www.biointelligence.hu/pyhubs.) library.
- 1.Fix, E., Hodges, J.L., Jr.: Discriminatory analysis-nonparametric discrimination: consistency properties. Technical Report, DTIC Document (1951)Google Scholar
- 8.Buza, K., Nanopoulos, A., Schmidt-Thieme, L.: Time-series classification based on individualised error prediction. In: 2010 IEEE 13th International Conference on Computational Science and Engineering (CSE), pp. 48–54. IEEE (2010)Google Scholar
- 9.Tomašev, N., Mladenić, D.: Nearest neighbor voting in high-dimensional data: Learning from past occurrences. In: 2011 IEEE 11th International Conference on Data Mining Workshops (ICDMW), pp. 1215–1218. IEEE (2011)Google Scholar
- 11.Bhattacharya, G., Ghosh, K., Chowdhury, A.S.: Test point specific k estimation for kNN classifier. In: 2014 22nd International Conference on Pattern Recognition (ICPR), pp. 1478–1483. IEEE (2014)Google Scholar
- 15.Tomašev, N., Buza, K., Marussy, K., Kis, P.B.: Hubness-aware classification, instance selection and feature construction: survey and extensions to time-series. In: Stańczyk, U., Jain, L.C. (eds.) Feature Selection for Data and Pattern Recognition. SCI, vol. 584, pp. 231–262. Springer, Heidelberg (2015). https://doi.org/10.1007/978-3-662-45620-0_11CrossRefGoogle Scholar