Jasper Gerigk
Jasper Gerigk
Home
Publications
Experience
Education
Projects
Contact
1
Learning Various Strategies For Dominion Using Deep Reinforcement Learning
We use a graph-based game representation and a modified Soft Actor-Critic algorithm to train a deck-building game agent that outperforms all previous learning-based approaches and manipulate training so that the agents exhibit novel human-like play strategies.
Jasper Gerigk
,
Steve Engels
PDF
Cite
Abstract
An Enhanced Graph Representation for Machine Learning Based Automatic Intersection Management
We expand on the previously developed graph-based scene representation and graph neural network to approach the automated intersection management problem using reinforcement learning and show that the model outperforms baselines and generalizes very well.
Marvin Klimke
,
Jasper Gerigk
,
Benjamin Völz
,
Michael Buchholz
PDF
Cite
Abstract
Cite
×