Angus Chadwick: Integrating the computation-through-dynamics and efficient coding frameworks: noise-robust and energetically efficient computation in neural circuits
| When |
Jul 22, 2026
from 12:15 AM to 01:00 PM |
|---|---|
| Where | Bernstein Center Freiburg, Hansastr. 9a, Lecture Hall, Ground Floor |
| Contact Name | Martina Bacher |
| Add event to calendar |
 vCal iCal
|
Abstract
In this talk, I will discuss recent work from our lab integrating the computation-through-dynamics and efficient coding frameworks. I will introduce a novel method for optimisation of noisy, continuous-time dynamical systems for efficient coding of task variables. I will then show how this approach yields previously overlooked solutions to working memory tasks involving high-dimensional rotational dynamics, which leverage properties of State Space Models from machine learning and account for diverse aspects of neural responses observed experimentally in prefrontal cortex and other brain areas.
Next, I will show how the same method can be extended to explain the organisation of sensory cortical circuits, providing a normative account of the circuit mechanisms underlying orientation tuning in visual cortex, including cell type-specific connectivity, contrast invariance and contextual modulation, and the reorganisation of tuning curves with perceptual learning. Taken together, these findings bridge mechanistic and normative accounts of neural circuits, providing a principled account of circuit-level phenomena across diverse brain regions and tasks.