SC²S Colloquium - July 20, 2017
|Date:||July 20, 2017|
|Time:||16:00 pm, s.t.|
Ivan Mauricio Rodriguez: Deep Reinforcement Learning for Superhuman Performance in Doom
Over the last years, Reinforcement Learning (RL) has attracted the attention of many researchers. Its powerful combination with Artificial Deep Neural Networks, when used as function approximators, has shown to be successful in many works. Rather than target classical RL problems, the most prominent examples of these works develop techniques that allow agents to learn how to play video and board games from raw input data at human level. In this thesis, we describe the implementation of an algorithm to train an agent for the popular 90's computer game Doom. Doom features several recurrent problems in RL such as delayed rewards and partial observability which are tackled by the algorithm. In particular, we discuss the efforts to improve the efficiency of our approach and the results obtained in several tests scenarios.
Andreas Schmelz: Optimization of Molecular Dynamics Simulations Using One-Sided MPI-Directives
Molecular dynamics simulations are an interesting and important field of research and application. An accurate simulation requires a lot of computing power. Therefore work is distributed between multiple machines in order to speed up simulation time. The MPI library with its matured point to point communication model is often used. A newer one-sided model has been introduced with MPI's latest releases. We will present MPI's two- and one-sided model, implement and optimize our simulation program with both models and compare them against each other using simulation scenarios as well as isolated micro benchmarks.