Joint exploration of learning systems

The motivation for the network is that to date a general understanding of learning systems and a comprehensive approach to their analysis and design is still largely missing. Learning systems are systems that can adjust and adapt their behavior to influences from their environment. Natural learning systems are, for example, the brain, the nervous system, and living organisms down to the smallest cells and bacteria. Artificial learning systems are, for example, robots that can adapt their behavior to their environments, or software systems making predictions based on internet data. Natural as well as artificial learning systems are often influenced by highly unreliable, stochastic factors. The main goal of the new research network is to achieve a fundamental understanding of the perception, learning, and adaptation in complex systems and to recreate characteristic properties of natural learning systems. The research network is also designed to address cross-disciplinary research questions in the design and analysis of natural and man-made learning systems.

Michael Black, the current managing director of the Max-Planck-Institute for Intelligent Systems, is very enthusiastic about the cooperation with his Swiss colleagues: “In this research network, two partners meet who are ideally suited for one another.” The outstanding basic science competence of the Max Planck Institute will be complemented by the excellent engineering science competences of the ETH Zurich.

 First projects already in progress

 An active exchange between the two institutions has already started. In addition to managing director Michael Black, the Max-Planck directors Bernhard Schölkopf, Stefan Schaal, Joachim Spatz and research group leader Peer Fischer contribute to the joined network from the Max-Plack-Institute for Intelligent Systems’ side. For example, scientists in the group of Professor Stefan Schaal from Tübingen are investigating together with Dr. Stelian Coros from the ETH new control and planning algorithms for locomotion on two legs for humanoid robots. They are trying to understand algorithms that can achieve robust and stable walking over very rough terrain. Together with Dr. Jonas Buchli from the ETH, this group also aims at tightly integrating multi-modal perception, motion planning and control to allow for a fluent and seamless coordination of sensing and acting in robots with arms and legs. Professor Joachim Spatz is working closely with Professor Viola Vogel-Scheidemann of the ETH to better understand the biophysics and biochemistry of how living cells sense their environment (mechanosensitivity) and navigate in tissues. These studies can lead to an improved understanding of the function and development of the immune systeme and diseases like cancer. Jointly with Dr. Balduzzi of the ETH computer science department, Prof. Schölkopf and his team have recently developed a novel method for assessing the strength of causal interactions between observables, to be published by the Annals of Statistics.

Another goal of the research network is to provide a platform for exchange in research and education to advance the training of young academics. It will be possible for Professors of the ETH and Directors of the MPI to jointly supervise PhD Students enrolled at the ETH Zurich. Students and young scientists will benefit from summer schools, and  special residence programs in which a group of around 20 scientists and students work together on a specific research topic. “With all this, we ensure that our young academics will have the chance of the most outstanding collaborations and exchange early on in their careers”, explains Stefan Schaal.

Michael Black is confident about the success of the newly formed collaboration: “This network is the start of a cooperation between two leading research institutions that will advance fundamental understanding of learning systems that adapt to complex environments. This network positions Europe as the leading place for research on intelligent systems worldwide.”