ByEnzo Tagliazucchi (Buenos Aires, Argentina).
In the last decades, machine learning (in particular, deep learning) has consolidated as a data-driven way to identify meaningful patterns in large volumes of data. In this talk I will explore the application of deep learning to the problem of identifying signatures of criticality and non-equilibrium dynamics in recordings of human brain activity. The key idea behind my talk is that a mathematical model known to exhibit critical behavior can be used to train a deep neural network with the purpose of identifying similar behavior in empirical data, offering an alternative to theory-driven and parametric data analysis, which can be highly challenging in the face of noisy or insufficient observations.