Phase Alignment Enhances Oscillatory Power in Neural Mass Models Optimized for Class Encoding

Abstract: Neural encoding of objects and cognitive states remains an elusive yet crucial aspect of brain function. While traditional feed-forward machine learning neural networks have enormous potential to encode information, modern architectures provide little insight into the brain's mechanisms. In this work, a Jansen and Rit neural mass model was constructed to encode different sets of inputs, aiming to understand how simple neural circuits can represent information. A genetic algorithm was used to optimize parameters that maximized the differences in responses to particular inputs. These differences in responses manifested as phase-shifted oscillations across the set of inputs. By delivering impulses of excitation synchronized with a particular phase-shifted oscillation, we demonstrated that the encoded phase could be decoded by measuring oscillatory power. These findings demonstrate the capability of neural dynamical circuits to encode and decode information through phase.