DeepCuriosity
General methods for curiosity-driven machine learning and their applications
Project : DeepCuriosity
Project leader : Pierre-Yves Oudeyer – équipe Flowers – Inria Bordeaux – Sud-Ouest
Applications : Autonomous agents in large open video games world and autonomous robot exploration, self-organized bio-printed cells systems, personalized curriculum of exercises for human learners in digital educational apps.
Aim: To go beyond both conceptual and practical limits of deep reinforcement learning, thanks to an approach grounded in modelling mechanisms of infant learning and development in humans, in particular curiosity-driven learning. To enable real world robots to learn efficiently repertoires of high-dimensional skills, as well as learning natural language interaction with other agents, while being able to adapt quickly to changes in the environment, and under limited time and energy resources.