Workshop

Deep Reinforcement Learning in Practice

Nima Siboni

Research Engineer, InstaDeep and Guest researcher, Max-Planck-Institute

Dr. Nima Siboni, machine learning team lead at DeepMetis, has set his focus on reinforcement learning projects. After finishing his Ph.D. in simulation sciences, he complemented his skills with machine learning techniques, and since then works as an AI/RL practitioner and researcher. His expertise lies at the intersection of AI and simulation, namely optimization, reinforcement learning, and simultelligence.

Nima Siboni
Nima Siboni
Session description

Unlike the un/supervised learning, which are extensively used in industry, RL is still not that often utilized, in spite of its potential. Nevertheless, there are recent developments in making RL easier to train and more reliable to use. In this workshop, we first examine for which problems RL is a suitable framework, and then we address a minimal but easily extendable use case that could serve as a blueprint for your business problem.


The outline of the tutorial goes as follows, covering from RL-basics to More advanced subjects; the audiences of the material are RL practitioners or those who want to consider it as a new approach to solve their problems.

- RL in AI ecosystem: What problems can you solve with RL?

-  It is not only for games: Successful RL in industry.

- RL essentials: Fundamentals of RL problem formulation.

- Industry-grade packages for RL

- Training and evaluating an RL agent

- (depending on time restrictions) Knots and bolts of RL: What if the agent is not learning?


  • Please install the following requirements:
    numpy~=1.22.4
    tensorflow~=2.9.1
    tensorboard~=2.9.1
    tqdm
    gym~=0.21.0
    ray~=1.12.1
    ray[rllib]
    pyglet~=1.5.26