LENGA

Behavioral adaptation relies on the selection of actions that maximize rewards and minimize costs by learning the relationships between actions and outcomes. When the opportunities to correct mistakes are limited to updating synapses, learning must occur instantaneously. LENGA explores novel learning algorithms where neural gain learning complements synaptic learning. By merging in vivo neurophysiology and behavior with computational models, LENGA will contribute to identifying biological details required to achieve more adaptive and generalizable learning for sustainable AI development. It involves collaboration with the Cophy Team (Centre de Recheche en Neurosciences de Lyon) and MathNeuro Team (Inria Branch at the University of Montpellier).