Stage-oe-small.jpg

Thema4584

Aus Aifbportal
Version vom 4. Mai 2021, 10:29 Uhr von Dh1659 (Diskussion | Beiträge)
(Unterschied) ← Nächstältere Version | Aktuelle Version (Unterschied) | Nächstjüngere Version → (Unterschied)
Wechseln zu:Navigation, Suche



Machine Learning applied to Robotic - Unsupervised Reinforcement Learning




Informationen zur Arbeit

Abschlussarbeitstyp: Master
Betreuer: Mohammd Karam Daaboul
Forschungsgruppe: Angewandte Technisch-Kognitive Systeme
Partner: FZI Forschungszentrum Informatik
Archivierungsnummer: 4584
Abschlussarbeitsstatus: In Bearbeitung
Beginn: 26. Januar 2021
Abgabe: unbekannt

Weitere Informationen

Reinforcement learning (RL) is a field of machine learning that deals with sequential decision-making aimed at maximizing a cumulative reward. Through the connection between RL and High-Capacity Representation like Neural Network, an AI-agent can learn to solve specific tasks from raw, low-level observations such as images. However, in continuous domains governed by complex dynamics, such as robotic control, it isn't easy to learn from these high-dimensional inputs directly. Standard model-free deep RL aims to use direct end-to-end training to unify these tasks of representation learning and task learning explicitly. However, solving both problems together is difficult, since an effective policy requires an adequate representation. The goal of this work is to use useful representations learned from the latent variable model and train effective RL agents in this learned latent space.