2000 character limit reached
FEMDA: Une méthode de classification robuste et flexible (2307.01954v1)
Published 4 Jul 2023 in stat.ML and cs.LG
Abstract: Linear and Quadratic Discriminant Analysis (LDA and QDA) are well-known classical methods but can heavily suffer from non-Gaussian distributions and/or contaminated datasets, mainly because of the underlying Gaussian assumption that is not robust. This paper studies the robustness to scale changes in the data of a new discriminant analysis technique where each data point is drawn by its own arbitrary Elliptically Symmetrical (ES) distribution and its own arbitrary scale parameter. Such a model allows for possibly very heterogeneous, independent but non-identically distributed samples. The new decision rule derived is simple, fast, and robust to scale changes in the data compared to other state-of-the-art method
- Carl J. Huberty. Discriminant Analysis. Review of Educational Research, 1975.
- Peter A. Lachenbruch. Discriminant Analysis When the Initial Samples Are Misclassified. Technometrics, 1996.
- P. A. Lachenbruch and M. Goldstein. Discriminant Analysis. Biometrics, 1979.
- Huber Peter J. Robust covariances. Statistical decision theory and related topics, 1977.
- Andrews Jeffrey L and McNicholas Paul D and Subedi Sanjeena. Model-based classification via mixtures of multivariate t-distributions. Computational Statistics & Data Analysis, 2011.
- Bose Smarajit and Pal Amita and SahaRay Rita and Nayak Jitadeepa. Generalized quadratic discriminant analysis. Pattern Recognition, 2015.
- Ghosh Abhik and SahaRay Rita and Chakrabarty Sayan and Bhadra Sayan. Robust generalised quadratic discriminant analysis. Pattern Recognition, 2021.
- Houdouin Pierre and Pascal Frédéric and Jonckheere Matthieu and Wang Andrew. Robust classification with flexible discriminant analysis in heterogeneous data. https://arxiv.org/abs/2201.02967, 2022.
- Roizman Violeta and Jonckheere Matthieu and Pascal Frédéric. A flexible EM-like clustering algorithm for noisy data. arXiv preprint arXiv:1907.01660, 2019.