[ABE-L] Ciclo de Seminarios PPGEst-UFRGS ( Anderson Ara - UFPR - 01/08 às 13:30)

[RENATA ROJAS GUERRA]

Prezados,

Temos o prazer de convidar a todos para mais uma palestra do Ciclo de
Seminários 2023 do Programa de Pós-Graduação em Estatística da UFRGS
(PPGEst-UFRGS)!

Informações abaixo:

*Palestrante: *Anderson Ara - UFPR

*Título:* Random Machines: A new machine learning method

*Data:* 01 de agosto de 2023 (terça-feira)

*Horário:* 13h30min às 14h45min

*Link:* mconf.ufrgs.br/webconf/ppgest

*Resumo:* Supervised machine learning techniques have one of their main
objectives to reduce the generalized prediction error. Support vector
models (SVM) have been drawing the attention of the community once these
models have some properties which are easy to characterize as well as allow
an estimation process with global optimization properties. However, SVM has
challenges in choosing the appropriate kernel function and the tuning
estimation of its hyperparameters. In this paper, we present a new machine
learning method, namely Random Machines (Ara et al., 2022). The proposed
method eliminates the need to choose the best kernel function during the
tuning process using a random mixture of kernel functions combined with a
properly new bagging procedure. In this paper, the application of the
Random Machines is illustrated by prediction of the general
psychopathological symptoms from the SIPS (Structured Interview for
Prodromal Syndromes) on At Risk Mental State (ARMS) individuals. The
covariates were extracted from facial movement of brief video recordings of
127 patients on medical appointments. As the main result, Random Machines
showed a superior predictive capacity than methods XGB (extreme gradient
boosting), RF (random forest), DMLP (deep multilayer perceptron), SVM and
LR (logistic regression). Project in partnership with LIM-27, Institute of
Psychiatry – USP, Wellcome Trust research funding.
-------------- Próxima Parte ----------
Um anexo em HTML foi limpo...
URL: <http://lists.ime.usp.br/pipermail/abe/attachments/20230801/b845d703/attachment-0001.htm>
-------------- Próxima Parte ----------
Um anexo não-texto foi limpo...
Nome: IMG-20230801-WA0018.jpg
Tipo: image/jpeg
Tamanho: 118851 bytes
Descrição: não disponível
URL: <http://lists.ime.usp.br/pipermail/abe/attachments/20230801/b845d703/attachment-0001.jpg>