You are here

Marta Ellero's graduation

News date: 
Thursday, 17 July, 2014

On July 17th, Marta Ellero successfully defended her Master in Cognitive Science thesis entitled "Inference with Pattern Classifiers in Neuroimaging Experiments". The focus of her thesis was methodological analysis of classifier use in brain decoding.

Recently, there has been increasing adoption of multivariate techniques from machine learning and pattern recognition communities for the analysis and inter-
pretation of fMRI data. Among these, a method that has become popular is pattern-based classification analysis. In pattern-based classification analysis, you test how
reliably a trained classifier can decode, i.e. predict, the experimental conditions that the subject is undertaking from distributed patterns of fMRI activity.

The author in her thesis shows the limits of this standard procedure with respect to both the metric used for the evaluation of the classifier performance and the framework of hypothesis testing adopted. Then, she presents a new solution to address the problem of statistical inference through classifiers on fMRI data.