Palestra: Statistical depth functions: Data science with theory.
Palestrante: Milan Merkle (University of Belgrade, Serbia)
Dando continuidade ao Ciclo de Palestras do Programa de Pós-Graduação em Estatística do IM-UFRJ, nesta 4a feira:
Data: 28 de agosto 2019
Horário: 15:30h
Local: Laboratório de Sistemas Estocásticos (LSE), sala I-044.
Resumo: About one year ago, David Donoho (professor of statistics at Stanford) gave a talk in Vienna University with the title Data Science-the end of theory. Just some months later, on the ICM in Rio de Janeiro, he was awarded the Gauss prize for his work on compressed sensing methods. This presentation is related to another Donoho's work, which is less known. The concept of a depth function was created by J. W. Tukey's attempts to define multidimensional counterparts to median and quantiles, in late 70's. While Tukey gave just an idea, it was Donoho who wrote the first theory of Tukey's depth function in his Ph.D. thesis. Thereafter, many other proposals for depth functions have appeared, but the Tukey's one is still most popular, although this concept does not belong to the mainstream methods in Data Science.