Olivier Bertrand is delighted to have Andreea Diaconescu from the Translational neuromodeling unit (University of Zurich) to come and give a talk about:
Inferring on hierarchical regularities: a computational perspective on the mismatch negativity component in the auditory system.
An emerging theme in neuroscience is the notion that the brain maintains and continuously updates a generative model of its environment, which allows for inference on the (hidden) causes of its sensory inputs (Doya et al., 2011; Friston, 2010). This mechanism was postulated to explain the generation of the mismatch negativity (MMN) potential, a noninvasive electrophysiological response to local violations of the regularity of sensory inputs (Garrido et al., 2008; Lieder et al., 2013). In this talk, I will use an information theoretic approach (Marr, 1982) to highlight some of the key mechanisms that give rise to the MMN potential in the auditory system at computational, algorithmic and implementational levels. Furthermore, I will extend this mathematical formulism to more complex cognitive processes such as learning about hierarchical (second-order) statistical regularities. Finally, I will address which aspects of these mechanisms may break in certain psychiatric disorders, such as schizophrenia and autism spectrum disorder (ASD).