The generation and dynamic control of neural activity sequences

by Andrew B. Lehr
Doctoral thesis
Date of Examination:2024-04-25
Date of issue:2025-02-24
Advisor:Prof. Dr. Christian Tetzlaff
Referee:Prof. Dr. Christian Tetzlaff
Referee:Prof. Dr. Stefan Klumpp
Persistent Address: http://dx.doi.org/10.53846/goediss-11100

 

 

Files in this item

Name:Lehr_PhD_Thesis.pdf
Size:48.5Mb
Format:PDF

The following license files are associated with this item:

Abstract

English

Our behavior and cognition is flexible and dynamic. With ease, we can switch between different thoughts, memories, and movements. We are able to dynamically adapt individual behaviors, like the speed and path our arm takes as we catch a ball. Even our experience itself is not rigid, but instead our brains pay more attention to and prioritize certain more important experiences over less important ones, influencing our decision-making and memory formation. Dynamic processes such as flexibly switching between behaviors, dynamically adapting a behavior, and prioritizing incoming information on the fly, all require fast timescale neural mechanisms for their implementation in neural circuits. In the central nervous system of animals, the firing patterns of neurons within local circuits form spatiotemporal sequences of activity. Thus underlying each of these behaviors, memories, or sensory experiences, is a spatiotemporal sequence in the relevant brain regions. And so, it stands to reason that the astounding flexibility we are capable of must be rooted in dynamic and controlled fluctuations in this underlying sequential neural code. In this thesis, we will investigate the neural circuitry and mechanisms underlying the generation and control of spatiotemporal activity sequences. We will extend a model of sequence generation in locally connected recurrent circuits based on observations from the cortex of animals. By considering neural mechanisms that act on a fast timescale, in a simplified network model we will be able to link dynamic changes in local properties, like neuronal gain, inhibition, synaptic currents and weights, to changes in sequential dynamics and their readout as neural trajectories on low dimensional subspaces. In particular, in recurrent networks for sequence generation, we will introduce mechanisms to store and flexibly switch between different behaviors, dynamically modulate behavior, and prioritize important sensory experiences to update our perspective and behavioral policies. This work is a step towards the development of a framework connecting neural mechanisms acting within local circuits to their effect on local activity dynamics and the low dimensional downstream readout. The goal being to develop an understanding of how the computations underlying flexible cognition and behavior are implemented within the interconnected networks of the central nervous system.
Keywords: computational neuroscience; spatiotemporal activity sequences; neural manifolds; flexible behavior; dynamic control
 
Statistik