Logo des Repositoriums

Auflistung nach Schlagwort "reinforcement learning"

1 - 3 von 3
  • Konferenzbeitrag
    Overview of machine learning and data-driven methods in agent-based modeling of energy markets
    (INFORMATIK 2019: 50 Jahre Gesellschaft für Informatik – Informatik für Gesellschaft, 2019) Prasanna, Ashreeta; Holzhauer, Sascha; Krebs, Friedrich
    Local energy markets (LEM) allow prosumers and consumers to trade energy directly between each other and offer flexibility services to the grid. The benefits and challenges related to such markets need to be identified, and agent-based modeling (ABM) is a useful method to conduct simulation experiments with different market structures and clearing mechanisms. Machine learning (ML) and data-driven methods when integrated with ABM show great potential for constructing new distributed, agent-level knowledge. In this paper, we discuss the requirements for coupling ML methods and ABM. We also provide an overview of published literature on the common methods of integration of ML and data-driven methods in ABM of energy markets and discuss how these requirements are commonly addressed.
  • Textdokument
    Reinforcement Learning-Controlled Mitigation of Volumetric DDoS Attacks
    (GI SICHERHEIT 2022, 2022) Heseding, Hauke
    This work introduces a novel approach to combine hierarchical heavy hitter algorithms with reinforcement learning to mitigate evolving volumetric distributed denial of service attacks. The goal is to alleviate the strain on the network infrastructure through early ingress filtering based on compact filter rule sets that are evaluated by fast ternary content-addressable memory. The reinforcement learning agents task is to maintain effectiveness of established filter rules even in dynamic traffic scenarios while preserving limited memory resources. Preliminary results based on synthesized traffic scenarios modelling dynamic attack patterns indicate the feasibility of our approach.
  • Textdokument
    Training a deep policy gradient-based neural network with asynchronous learners on a simulated robotic problem
    (INFORMATIK 2017, 2017) Lötzsch, Winfried; Vitay, Julien; Hamker, Fred
    Recent advances in deep reinforcement learning methods have attracted a lot of attention, because of their ability to use raw signals such as video streams as inputs, instead of pre-processed state variables. However, the most popular methods (value-based methods, e.g. deep Q-networks) focus on discrete action spaces (e.g. the left/right buttons), while realistic robotic applications usually require a continuous action space (for example the joint space). Policy gradient methods, such as stochastic policy gradient or deep deterministic policy gradient, propose to overcome this problem by allowing continuous action spaces. Despite their promises, they suffer from long training times as they need huge numbers of interactions to converge. In this paper, we investigate in how far a recent asynchronously parallel actor-critic approach, initially proposed to speed up discrete RL algorithms, could be used for the continuous control of robotic arms. We demonstrate the capabilities of this end-to-end learning algorithm on a simulated 2 degrees-of-freedom robotic arm and discuss its applications to more realistic scenarios.