This part contains some of my academic work. Most notably my bachelor and master thesis and my published papers. I translated my theses from LaTeX code to reStructuredText, so there may be some minor differences between the PDF and HTML version. The PDF documents are the complete and submitted versions.

## Master thesis¶

My masterthesis shows that interactive feedback leading to revert an action improves the learning speed of reaching with a planar two Degrees of Freedom robotic arm. I found no advantage in interactive feedback with one Degree of Freedom.

You can also download the thesis and find the source code here.

## Bachelor thesis¶

In my bachelorthesis I modified the class diagram within BlueJ so that the user can change the visualisation of a class from the implementation, or standard view, and interface view in which only the externally visible dependencies are visible; all dependencies that only occur in the implementation and not in the interface are hidden from view.

I wrote this thesis in German and you can find it here.

## Publications¶

Interaction in reinforcement learning reduces the need for finely tuned hyperparameters in complex tasks

Kognitive Systeme, Nr. 3 (2016). doi:http://dx.doi.org/10.17185/duepublico/40718. Stahlhut, Chris, and Navarro-Guerrero, Nicolás, and Weber, Cornelius, and Wermter Stefan. PDF, BibTeX

Dynamische Klassendiagramme-Nutzung Der Metapher Vom ‘Konsumieren Und Produzieren’ in BlueJ

In SEUH, 21–26, 2011. Schmolitzky, Axel, and Stahlhut, Chris PDF, BibTeX