Below you can find a list of projects in which we participate:

Advanced Methods for Automated Optimization and Modeling of the Empirical Performance of Highly Parameterized Heuristic Algorithms

Project page (in german)

Auto-Tune: Structural optimization of Machine Learning frameworks for large datasets

As part of the DFG Schwerpunktprogramm 1527 on "Autonomous Learning", we work on algorithm selection and algorithm configuration for machine learning. One of the reasons why machine learning is currently underused is the limitation of tools that require little to no experience from the end user. The recent AutoWEKA approach has demonstrated the possibility to simultaneously choose a feature selection strategy, a learning algorithm and its hyperparameters in the to optimize empirical performance automatically. While Auto-WEKA provides an effective way for non-experts to achieve state-of-the-art performance in terms of human time spent, it can be costly in terms of machine time, as it requires the evaluation of thousands of configurations to reliably home in on good configurations. These computational demands can be tolerated for modestly-sized data sets, and for model classes that are inexpensive to train, but they are unacceptable for large data sets and/or computationally expensive models, such as deep belief networks and convolutional networks. The main goal of this project is to overcome these limitations and make the automatic instantiation of complex model families practical for large datasets.

Project page


The RobDREAM project aims to improve the performance of robots for tasks involving perception, navigation, grasping and mobile manipulation. Inspired by learning in humans,the idea is simple: while the robot works, it records data about its tasks and the achieved performances. After a workday, in the nightly rest phase, data from this and earlier days is analyzed. In the same way sleep is important for brain development in humans, this “dreaming” is supposed to be an integral part of the learning phase the robot undergoes while mastering a new task. In this EU project, we collaborate with partners from industry and academia alike. The other members of the consortium are experts in their respective fields of robotics, and we contribute with our expertise in algorithm configuration.

Project page


The constantly growing amount and complexity of data involved in brain studies makes manual parameterization of data processing and modelling algorithms a tedious and time-consuming task. This can preclude some BrainLinks-BrainTools projects from reaching their full potential due to the use of poorly parameterized tools. This project aims to provide tools for automated hyperparameter tuning to nable continuous exploration of capabilities and improvements within BrainLinks-BrainTools projects. The special focus is on automated deep learning representations aimed at extracting discriminative patterns from brain signals that can be used as input for machine learning methods.

Project page