GRK 1589:  Sensory Computation in Neural Systems

• Subject Area: Neurosciences
• Term: from 2010 to 2019
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 103586207

In the Research Training Group techniques and concepts from machine learning, computational neuroscience and system neurobiology are combined in order to specifically address sensory computation. Experimentalists and theoreticians will join forces and educate young scientists (1) to work on interdisciplinary projects investigating brain computation (rather than phenomena related to neural response properties or neural dynamics only), (2) to exploit recent advances in machine learning, theoretical computer science and statistics for modelling brain function, and (3) to develop new theories of computation hand in hand with well-controlled experiments in order to put functional hypotheses to test.
Our graduates will be competent in machine learning, computational neuroscience, cognitive science and neurobiology, and will be able to judge scope and limits of theoretical and experimental approaches - in the ideal case conducting the proper experiments themselves. The scientific programme is focussed on sensory coding and perception. Since perception is task dependent (“perception serves a purpose”) sensory processing is connected to cognitive functions (decision making, memory function, planning and even motor control) and has to be linked to performance measures.
Hence, all experiments will involve behaving animals or human subjects. Exploiting new ideas from the machine learning field, theoretical concepts will be developed for specifically addressing temporally varying inputs, coding strategies for stimulus time series and computation with dynamical systems. Experiment-theory projects are conducted by at least one investigator with a strong experimental and one investigator with a strong theoretical background.
Hence, the four key aspects of this Research Training Group are: (1) its focus on functional hypotheses about perception (“computation” rather than “dynamics”), (2) its focus on behavioural paradigms, often linking perception with cognitive functions, (3) its focus on advanced machine learning techniques not yet widely used for the analysis and modelling of neural systems and (4) its focus on interdisciplinary training implemented by strict co-supervision through researchers with complementary expertise.

• DFG Programme: Research Training Groups
• Applicant Institution: Technische Universität Berlin
• Co-Applicant Institution: Freie Universität Berlin; Humboldt-Universität zu Berlin
• Spokesperson: Professor Dr. Klaus Obermayer
• Participating Researchers: Professor Dr. Felix Blankenburg; Professor Dr. Michael Brecht; Professor Dr. John-Dylan Haynes; Professor Dr. Andreas Heinz; Professor Dr. Richard Kempter; Professor Dr. Benjamin Lindner; Professor Dr. Klaus-Robert Müller; Professor Dr. Manfred Opper; Professorin Dr. Susanne Schreiber; Professor Dr. Imre Vida