2024-07-10 –, If (1.1)
Today, neuroscience is generating data more than ever before. A central challenge is how to effectively leverage this unprecented data deluge while embracing existing and emerging domain knowedge. We discuss how emerging developments happening in scientific machine learning present untapped opportunities to advance the field. To harness this potential, we present the first version of NeuroDynamics.jl which aims to provide the building blocks towards next generation modelling in neuroscience.
While the causes of single neuron spikes have been understood for decades, the processes that support collective neural behavior in large-scale cortical systems are less clear. Understanding such large-scale processes is essential to uncover the neural basis underlying cognitive and behavioral functions and dysfunction. Thus, modeling neural dynamics across different scales of organization and levels of abstraction has been an area of great interest in neuroscience.
In our work, we aim to develop a framework to enable the integration of different modeling techniques under a common language. NeuroDynamics.jl provides a suite of well-established and emerging models used in neuroscience. These range from biophysical, phenomenological, data-driven, and, most importantly, hybrid models. These hybrid models capitalize on recent developments in the scientific machine learning community to tackle long-standing challenges that exist with current models. We provide tutorials to show how these hybrid models can be used to understand the neural computation underlying cognitive and behavioral tasks, system identification for neural control, neural decoding, as well as normative modeling of neural function.
The long-term vision of NeuroDynamics.jl is to be a one-stop shop for modeling, simulation, analysis, and control of any neural system.
Ahmed is a postdoctoral reseacher at the donders institute for brain cognition and behaviour. His research focuses on modelling and control of complex adaptive systems.