My journal, my point of view, my journey with its ups, downs and full disclosure.
-
From Georgia Tech AutoRally to SNOW-AutoRally ... and beyond
Research internship oral presentation at the Northern Robotics Laboratory (Norlab) of Université Laval on mobile robotic in adverse condition and the Information-Theoretic Model Predictive Control algorithm. (~1 hour 30 min)
-
Une intuition sur RUDDER
Présentation de l'article RUDDER: Return Decomposition for Delayed Rewards écrit par Arjona-Medina, J. A. et al. dans le cadre du cours GLO-7030 Apprentissage par réseaux de neurones profonds donné à l'Université Laval. (~6 min)
-
Soft Actor-Critic part 1: intuition and theoretical aspect
How to teach robustness to a deep reinforcement learning agent using the maximum entropy principle. In this essay, I cover the building blocks of the SAC algorithm and the relevant nuts and bolts of the Maximum Entropy RL framework.
-
Do implementation details matter in Deep Reinforcement Learning?
A reflection on design, architecture and implementation of DRL algorithms from a software engineering perspective applied to research. Spoiler alert … it does matter!