NeuroAI

Numerous tools employed in the domain of artificial intelligence (AI) have drawn inspiration from insights obtained from neuroscientific research. We seek to further bridge AI methodologies with recent advances in human neuroscience, primarily by incorporating temporal dynamics into Deep Neural Networks. By constraining their dynamics and architecture through human MEG data, we intend to integrate temporal dynamics, including repetition suppression, facilitation, and neuronal oscillations, into these networks. Our primary objective is to develop models for visual processing in the human ventral system that are inspired by unsupervised learning algorithms. Additionally, we aspire to inspire enhancements in AI learning algorithms for visual processing.