In his Distinguished lecture, Prof. Górriz describes a brief story of the application of machine learning to neuroscience and describes several methodologies from the neuroimaging datasets to the assumptions and ubiquitous models employed for neuroimaging data analysis.
In the last decades, neuroscience has transitioned from reports of case studies to population studies. In this scenario, we usually utilize classical statistics, but it is only relatively recently that statistical learning methods including machine learning enjoy increasing popularity in this field. Despite the culture collision that happened from that moment, there’s a shift from data models to model-free approaches from their imaging data analysis that is improving our understanding of the human brain function.
The first use of machine learning methods dates from the early 90s, e.g principal component analysis to identifying subgroups. At the beginning of the 21st century, these methods became popular after renaming them as mind-breathing brain recovering or multivariate pattern analysis. In 2006 SiPBA applied for its very first research project “DENCLASES” (Neurological Disease Detection based on Signal Classification and BSS). However, there is an open question about the usefulness and interpretation of machine learning methods, when they are applied to simple methods for solving binary classification problems.
Prof. Górriz explores the complete connection between the univariate GLM and MLE regressions and derives a refined statistical test with the GLM based on the parameters obtained by a linear Support Vector Regression (SVR-iGLM).
Juan Manuel Górriz is currently a Full Professor with the Data Science and Computational Intelligence (DASCI) Institute at the University of Granada, Spain, a member of Cambridge NeuroScience, UK, and a member of the Ellis Network, EU. His research interests lie in the field of statistical signal processing in biomedical applications.