The team “Brain Networks and Learning” (BraiNets) is a research team working at the interface between system/cognitive neuroscience and artificial intelligence. BraiNets exploits most advanced theoretical approaches and computational tools from artificial intelligence and statistics to address open questions in cognitive and system neuroscience. In particular, the main objective of Brainets’ research is to better understand both the neural and computational bases of learning. At the neural level, we aim at clarifying the role of cortico-cortical interactions in learning and goal-directed behaviors. At the computational level, we seek to develop innovative computational models inspired from artificial intelligence and anchored on principles of biological systems. Our objective is to map the predicted learning computations onto cortico-cortical interactions and brain network dynamics.
Computational and theoretical approaches
The main theoretical approaches and methodology characterizing BraiNets are:
- Functional Connectivity analysis and information theory
- Computational neurophysiological and analysis of brain network data (MEG, intracranial EEG, LFP and fMRI)
- Bayesian probabilistic modeling and reinforcement learning
- Neural networks models and neurocomputational modeling
- Machine learning
- CausaL (ANR) : Neural and computational bases of human goal-directed causal learning by combining human neurophysiology (MEG and intracranial SEEG) and neuroimaging (fMRI) techniques with computational models of learning, such as Reinforcement Learning and Active Inference.
- NetScovery (HBP, EBRAINS): Hybrid approaches combining data-driven and model-based approaches for the study of cognitive functional connectivity and the validation of whole-brain models .
- AgileNeuroBot (ANR): Neuromorphic systems for integrated bio-inspired rapid visual detection and stabilization dynamics, and the development of robots to fly independently, without any user intervention.