Na próxima quarta-feira, 02 de julho, às 10 horas, na sala 2077 do Instituto de Ciências Exatas (ICEx), haverá a palestra “Machine Learning for Uncertain Data in the Life Sciences”, com o professor Alex Freitas, da University of Kent, UK. O evento é aberto ao público e não há necessidade de inscrição prévia.
Abstract
Standard machine learning algorithms were not designed to cope with data uncertainty, where the values of predictive features are expressed as a probability distribution, rather than a certain value. In this talk we propose some approaches to adapt classification (supervised learning) algorithms to cope with uncertainty in feature values. We investigate this kind of adaptation in two types of ensembles for classification, random forests and ensembles of Naïve Bayes classifiers. We evaluate the proposed approaches on two application domains in the life sciences: predicting whether a gene has a pro-longevity or anti-longevity effect on a model organism (e.g. mice), and predicting whether or not a drug has a certain side effect. We report results for both types of ensembles in terms of predictive performance, and we also discuss how to interpret (or explain) the predictions of one of the best Naïve Bayes ensembles by using two different feature importance measures, a standard measure and a new measure.
Saiba mais sobre o Alex no site: https://www.kent.ac.uk/school-of-computing/people/3057/freitas-alex
