Vortrag von Michael Wibral (Göttingen): "Using information theory to test predictive coding theories."

Wann 10.12.2019
von 18:00 bis 20:00
Termin übernehmen vCal

Affective Neuroscience and Psychophysiology  



Michael Wibral


10.12.2019 18:00 - 20:00 — GEMI, Raum 1.134, Gosslerstr. 14, Goettingen


Using information theory to test predictive coding theories

Predictive coding as a theory arguably dominates today's scientific discourse on how the cortex works. Importantly, it is positioned as a theory of general cortical function – yet, empirical tests so far are limited to situations with an explicitly predictive experimental context, simply to allow for a meaningful analysis. In other words, to find and understand the neurophysiological correlates of predictions and errors, experiments posit a priori, when and what is being predicted in which brain region. There are two problems with this approach: first, knowing what is being predicted when and where in the brain seems to require already a fair understanding of how the brain, or the cortex, works – which may not generally be the case yet. Second, restricting empirical tests of a general theory to experimental contexts that are explicitly designed with predictions in mind, in a strict sense prohibits conclusions about the applicability of that theory in other contexts. One might provocatively frame this problems as: "Is the cortex doing predictive coding when we don't test it?". Last but not least, this restriction excludes drawing on the vast majority of empirical neurophysiological data for testing (and possibly refuting) the theory. i.e. all data that were obtained with a focus on other descriptions of cortical function(s). In this talk I will introduce predictive coding theory along with an information theoretic framework for testing predictive coding theories by translating the concepts of predictability, predictions and errors (surprise) into information theoretic quantities measurable from data.