Embedded chimera states in recurrent neural networks

Abstract Fully and partially synchronized brain activity plays a key role in normal cognition some neurological disorders, such as epilepsy. However, the mechanism by which synchrony asynchrony co-exist population of neurons remains elusive. Chimera states, where coexist, have been documented only for precisely specified connectivity network topologies. Here, we demonstrate how chimeras can emerge recurrent neural networks training to display with machine learning. These solutions, refer embedded chimeras, are generically produced matrices slightly perturbed from random networks. We also that learning is robust different biological constraints, excitatory/inhibitory classification (Dale’s law), sparsity connections circuits. The be trained switch chimera solutions: an input pulse trigger unsynchronized groups chimera, reminiscent uni-hemispheric sleep variety animals. Our results imply emergence quite generic at meso- macroscale suggesting their general relevance neuroscience.