Bc. Karol Kuna

Master's thesis

Hluboké posilované učení s modelem prostředí a spojitými akcemi

Model-Based Deep Reinforcement Learning with Continuous Actions
Abstract:
Táto práca študuje využitie modelu prostredia v oblasti hlbokého učenia posilňovaním so spojitými akciami, kde tradičné metódy model prostredia nepoužívajú. Súčasťou práce je teoretický popis nového algoritmu, nazvaného „Deep Model Learning Actor-Critic“, ktorý porovnávame s existujúcou metódou „Deep Deterministic Policy Gradient“. Tieto metódy porovnávame z hľadiska schopnosti riešiť nové úlohy a …more
Abstract:
In this thesis, we study the application of an environment model to deep reinforcement learning with continuous actions, where contemporary methods are typically model-free. We give a theoretical description of a novel model-based actor-critic deep reinforcement learning technique that we developed, called Deep Model Learning Actor-Critic. We compare it with a model-free method, Deep Deterministic …more
 
 
Language used: English
Date on which the thesis was submitted / produced: 22. 5. 2017

Thesis defence

  • Date of defence: 19. 6. 2017
  • Supervisor: doc. RNDr. Tomáš Brázdil, Ph.D.
  • Reader: RNDr. Vojtěch Řehák, Ph.D.

Citation record

Full text of thesis

Contents of on-line thesis archive
Published in Theses:
  • světu
Other ways of accessing the text
Institution archiving the thesis and making it accessible: Masarykova univerzita, Fakulta informatiky

Masaryk University

Faculty of Informatics

Master programme / field:
Informatics / Artificial Intelligence and Natural Language Processing