Date: Wednesday, 06.11.2024 15:20-17:00 CET
Location: Building S1|15 Room 133
Abstract:
Neural activities in the cortex tend to be much richer — and thus frustratingly more complex — than our typical computational models make them seem. In this talk, I will present two different projects that aim to explain some of this complexity from first principles. First, I will will present a model of sampling-based probabilistic inference in primary visual cortex that explains a suit of ubiquitous cortical phenomena which had lacked a unifying explanation: noise variability, oscillations (in the gamma band), and stimulus-induced transients.
Second, I will present a theory of optimal information loading into working memory that challenges deeply entrenched intuitions about the operation of attractor networks. As it turns out, this theory explains the puzzling but robustly observed phenomenon of dynamic coding in prefrontal cortex. Together, these studies reveal a shared principle underlying cortical dynamics: seemingly idiosyncratic dynamical motives may serve well-defined computational functions.
