Integrating Biology and AI: Enhancing Recurrent Networks of Spiking Neurons for Visual Information Processing

  In this talk, we will explore the remarkable computing capabilities of the brain, focusing on recurrent networks of spiking neurons (RSNNs). Although RSNNs have been finely tuned by evolution, experience, and natural selection, the learning processes behind this optimization remain largely unknown. By leveraging advanced machine learning techniques, especially backpropagation through time (BPTT), we have created RSNN models that emulate the functional properties of their biological counterparts. Integrating Neuropixels experimental data with deep learning optimization tools, we present a biologically realistic model of the mouse primary visual cortex (V1) and lateromedial visual area (LM). This robust computational framework offers profound insights into visual information processing, effectively bridging the gap between artificial and biological neural networks.