How does machine learning help with internships

Machine learning, artificial intelligence, and business intelligence internship

The Bachelor / Master internship "Machine Learning", "Artificial Intelligence", and "Business Intelligence" of the Institute for Machine Learning and Information Systems (ISMLL) of Prof. Schmidt-Thieme, which takes place every semester with different topics, offers students the opportunity to become familiar with to deal with current issues from the areas mentioned.

The topics are always based on the institute's events in the previous semester.

In the summer semester 2010 a contribution to the Robocode project was made as part of the internship "Machine Learning and Artificial Intelligence". Robocode is learning software that puts the basics of machine learning in a practical context and thus makes it easier to get started with this subject. The aim of the software is to implement a so-called "bot" that competes in tournaments against bots from other participants. The original software was written in Java, but has already been transferred to the .NET Framework and used.

The aim of our implemented bot was to record battles in advance, analyze them and later use the knowledge gained. Specifically, log files were created and evaluated using time series analysis methods (feature extraction, PCA, SAX). The movement patterns obtained from this were assigned to different classes of bots and loaded into the current bot as a finished model. These models are used to classify the current opponent and employ an adapted counter strategy.

The bot was implemented in Java. Other languages ​​(Phyton) and technologies / software (Weka) were also used to analyze the time series. The computing cluster of the ISMLL was made available for complex calculations.

The project was carried out by Carsten Witzke under the supervision of André Busche.