NeuroMeetup June: How to navigate in an uncertain world ?
Time & Location
About The Event
First speaker: Anna Kutschireiter, PhD (Jan Drugowitsch’s lab, Harvard University)
Second speaker: Melanie Basnak (Rachel Wilson’s lab, Harvard University)
Abstract 1: The fly’s internal compass - a Bayesian perspective
Successfully navigating the world requires animals to keep track of their orientation by an ‘internal compass’. Some animals’ behavior suggests that they additionally track uncertainties in their orientation, and make strategic use of these uncertainties to adapt their behavior. The strategic use of uncertainties in both computation and behavior is a hallmark of dynamic Bayesian inference, but how such inference is implemented in the brain remains unknown. Here, we propose a recurrent neural network model that suggests such an implementation. It represents its heading direction estimation together with the estimate’s uncertainty as a bump of neural activity in a ring-structured network, with a topology that matches the Drosophila connectome. This thus demonstrates that the Drosophila HD system could in principle implement Bayesian HD tracking.
Bio: Anna is a postdoctoral research fellow in Jan Drugowitsch’s lab at Harvard University. She is interested in developing models of dynamic information processing in the brain, and currently focusses on heading direction tracking in the fruit fly. Before joining the Drugowitsch lab, she worked with Jean-Pascal Pfister on the theory of dynamic Bayesian inference and its application to problems that arise in neuroscience, first as a PhD student in Zurich, and later as a postdoc in Bern.
Abstract 2: Uncertainty representation and uncertainty-driven learning in the fly’s navigation system
Uncertainty is an inherent component of neural computation: it is present at every stage, from sensory perception to motor control. There is evidence to suggest that different animals, from insects to humans, make use of uncertainty to shape their behavioral choices. However, there is little evidence on how this might occur at a neural level. Here, we make use of the genetic and connectivity tools in the fruit fly Drosophila melanogaster to study this problem. We are performing calcium imaging experiments to track the activity of a subset of neurons, which are known to track the fly’s heading direction in a compass-like manner. The flies are walking on an air-supported ball, surrounded by an LED arena where we can present visual stimuli that are coupled or uncoupled to the animal’s movements. The experiments we have conducted so far suggest that varying the reliability of the visual cues changes the cue representation in the compass neurons, and that this may affect visual learning in the system.
Bio: Melanie Basnak is originally from Buenos Aires, Argentina, where she studied Biology at the Universidad de Buenos Aires, specializing in Animal Physiology and Neuroscience. As an undergraduate, she researched how polarized light is detected by crabs, conducting behavior and electrophysiology experiments. She is currently a fourth-year graduate student in the PhD Program in Neuroscience at Harvard University, working in Rachel Wilson’s lab. In the Wilson lab, she studies the fly’s compass system. She focuses on how uncertainty is represented in the brain, and how it affects learning.