Search

Contains fulltext : 293084.pdf (Publisher’s version ) (Open Access) The brain is a computational device that detects stimuli from the external world and directs behavioral responses. Accurate computations are, therefore, crucial for the brain. Unlike computers, which use digital signals for their computation, the brain uses analog signals to process information. Also, neurons, the basic I/O units in the brain, are far more complex than the logic gates used in computers. As a result, neural activity involves significant variability. The enriched external world also makes the sensory inputs highly diverse, further increasing neural systems' variability. Starting from this point, we discuss how the nervous system maintains certain invariances under rich variability to achieve stable and accurate computations. We use three specific research projects to illustrate this question from different perspectives. In the first project we propose an efficient neuronal coding theory. We calculate how a group of neurons could adjust their tuning curves to reduce the influence of noise or biophysical constraints and keep the information invariant. In the second project, we investigate the response of the Drosophila peripheral visual system to contrast stimuli. A luminance-sensitive neuron type is indispensable for maintaining invariant contrast perception. Finally, in the third project, we build a neural network model to simulate the generation and evolution of sequential activity in the cerebral cortex under synaptic plasticity. We find that homeostatic plasticity mechanisms that maintain the invariances of overall connectivity and neural activity are necessary to generate sequential activity. These three aspects of research correspond to the three levels of analysis for information processing: computation, algorithm, and implementation, constituting an overall understanding of neural computation. It is important to note that we place an emphasis on understanding neural computation at the neuronal level, which is evident in all of our research projects. Radboud University, 14 juni 2023 Promotores : Helmstädter, M.N., Gjorgjieva, Julijana 307 p.

Contains fulltext : 226301.pdf (Publisher’s version ) (Open Access) Radboud University, 24 november 2020 Promotor : Helmstädter, M.N. Co-promotor : Letzkus, M. 133 p.