A1406
Title: A Bayesian nonparametric approach to discriminant analysis
Authors: Laura D Angelo - Universita di Milano Bicocca (Italy)
Bernardo Nipoti - University of Milan Bicocca (Italy) [presenting]
Tommaso Rigon - University of Milano-Bicocca (Italy)
Abstract: A Bayesian nonparametric framework is introduced to improve classical discriminant analysis, particularly in scenarios with sparse data. The method provides a flexible approach that encompasses both linear and quadratic discriminant analysis as special cases. The key innovation lies in allowing information sharing across groups to improve the estimation of group-specific covariance matrices. This is accomplished through a scale-only nonparametric mixture model defined on the space of positive definite matrices. The use of a conjugate nonparametric prior ensures tractability and remarkable ease of implementation. Applications to both simulated and real datasets demonstrate the adaptability and effectiveness of the proposed methodology.
