Open Conference Systems, 50th Scientific meeting of the Italian Statistical Society

Font Size: 
Nonparametric penalized likelihood for density estimation
Federico Ferraccioli

Last modified: 2018-05-10


In this work we consider a nonparametric likelihood approach to multivariate density estimation with a regularization based on the Laplace operator. The complexity of the estimation problem is tackled by means of a finite element formulation, that allows great flexibility and computational tractability. The model is suitable for any type of bounded planar domain and can be generalized to the non-Euclidean settings. Within this framework, we as well discuss a new approach to clustering based on the concept of diffusion in a potential field, and a permutation-based procedure for one and two samples hypothesis testing.

Full Text: PDF