Research

My research interests are the followings:
  • Reinforcement Learning: This area of Machine Learning (ML) aims at solving large Markov Decision Processes (MDP) where the dynamics is only known through interactions with the dynamical system modeled as an MDP. The aim is to be able to control large and complex dynamical systems wtihout the need to establish a model.
  • Learning from Demonstrations: It consists in training an apprentice agent to realize a specific task. To do so, the apprentice observes an expert agent that realizes the task and has to learn form those observations how to do as well as the expert. I study two different ways to tackle this problem which are Inverse Reinforcement Learning (IRL) and Imitation Learning (IL).
  • Inverse Reinforcement Learning: It consists in finding the reward function that could explain the expert behavior. The apprentice would have to optimize this reward in order to obtain a behavior as good as the one of the expert.
  • Imitation Learning: It consists in directly imitating the expert policy. The tools are the ones used in Supervised Learning such as Classification algorithms. The drawback of using pure classification algorithm is not taking into account the underlying dynamics of the MDP. That is why, we are interested in creating classification algorithms that take into account this dynamics.
  • Game Theory: It consists in learning strategies to play well against any adversary in a given game.