Nicolas Perrin-Gilbert
Date:
Speaker
Nicolas Perrin-Gilbert has been a CNRS researcher at ISIR since 2013. His scientific aim is to develop effective learning algorithms that can address various challenges encountered in robotics, with a specific emphasis on reinforcement learning (RL) and applications related to balance control, bipedal locomotion and robotic walking aids.
Abstract
The talk will present Yomix, an interactive tool designed to explore low-dimensional embeddings of omics data, along with related work carried out in collaboration between ISIR and the Curie Institute. On the robotics and reinforcement learning side, the talk will introduce O3P, a library for off-policy and offline RL that is currently being developed. It will also highlight AFU, a recent method enabling online RL with an algorithmic approach that, until now, had proven effective only in offline settings.