Multivariate Distributions
Our main goal is to propose multivariate models for non Gaussian data. The method employed involves quadratic forms based on deviance residuals, similarly to the multivariate Gaussian distribution. Between the results, we have the assimilation of the behaviour of the distributions in relation to its parameters and the computational implementation.
Extensions to Bayesian Regression Trees
Extensions to Bayesian Regression Trees.
nlmSet
Nonlinear regression models catalogue. The goal of this project is develop a R package with a Shiny interface to document and explore nonlinear regression models. The model curve can be manipulated using sliders, so allowing the user select the most apropriate model for applications.Team: Augusto Buzzoni Calcanhoto, Willian Ramos & Bruna Wundervald.
Regularization for Tree-Based Models
We generalise the approach of Deng & Runger(2012), which relies on gain penalisation for fea-ture selection in Random Forests. We show that previous methods do not perform sufficient regularization and often exhibit sub-optimal out-of-sample performance, especially when correlatedfeatures are present. Instead, we extend the gain penalisation idea to a general local-global regularization for tree-based models, that allows formore flexibility in the choice of feature-specific importance weights.
R-Ladies
As a diversity initiative, the mission of R-Ladies is to achieve proportionate representation by encouraging, inspiring, and empowering people of genders currently underrepresented in the R community. R-Ladies’ primary focus, therefore, is on supporting minority gender R enthusiasts to achieve their programming potential, by building a collaborative global network of R leaders, mentors, learners, and developers to facilitate individual and collective progress worldwide.