Nesta quinta-feira, o Prof. Álvaro Veiga fala sobre sua pesquisa intitulada “Boosting smooth transition decision trees for partial effect estimation”

A apresentação será às 14h desta quinta-feira, 23 de agosto, na sala Multimeios do Departamento de Engenharia Elétrica da PUC-Rio*

Segue um pequeno resumo, disponível apenas em inglês, sobre o teor da palestra:

"In the lectue, we will introduce a new machine learning model for nonlinear regression called Boosting Smooth Transition regression tree (BooST). The model assumes a smooth predictive surface, allowing for the calculation of partial effects and the growth of smaller trees. In our formulation, we use a boosting algorithm to construct an additive nonparametric regression model consisting of a weighted sum of smooth transition regression trees.

"We will cover the basic concepts of Random Forests and present some fresh results showing its superior performance in predicting the US inflation in a data rich context. Next, we will describe the main elements of smooth transition trees and present our implementation of the gradient boosting algorithm to construct the predictive surface. We also provide some asymptotic theory showing the consistency of the estimated partial derivatives and present applications on simulated and real data."

*DEE – Departamento de Engenharia Elétrica
Rua Marquês de São Vicente, 225
Edifício Cardeal Leme, sala 401 – Gávea