Research focuses on significant improvements of performance and accuracy in application specific computing through a global optimization across the entire spectrum of numerical methods, algorithm design, software implementation and hardware acceleration.
These layers typically have contradictory requirements and their integration poses many challenges. For example, numerically superior methods expose little parallelism, bandwidth efficient algorithms convolve the processing of space and time into unmanageable software patterns, high level language abstractions create data layout and composition barriers, and high performance on today's hardware poses strict requirements on parallel execution and data access. High performance and accuracy for the entire application can only be achieved by balancing these requirements across all layers.
The following topics are given particular attention: Mixed precision methods, Multigrid methods, Adaptive data structures, Data representation, Bandwidth optimization, Reconfigurable computing.
Computing architectures which go beyond traditional von-Neumann concepts are needed to satisfy ever-increasing processing demand in artificial intelligence, advanced data processing and post-quantum security. We develop computing systems inspired by the structure of the brain and by recent advances in quantum physics, to develop massively parallel and low-power solutions to the most demanding applications.
We have developed the BrainScaleS accelerated analog neuromorphic system. Based on CMOS microelectronics technology, BrainScaleS is a flexibly configurable analog computing system. It provides a platform for experimental research into novel computing paradigms based on analog computing. Possible applications range from brain emulation, providing event-based (“spikes”) communication and structured neurons, to analog linear algebra and convolutional deep neural networks.
In addition to developing novel hardware concepts, we also focus on the software support to integrate analog computing in today’s ubiquitous digital infrastructure. BrainScaleS is part of the upcoming European EBRAINS research infrastructure for digital neuroscience.
We design integrated photonic circuits for neuromorphic and quantum computing and nanofabricate chipscale systems in our cleanrooms. Both integrated optical chips and control periphery are manufactured in house, making use of precision nanoprocessing and nanoanalytics. Photonic computing approaches offer massive gains in throughput and processing speed, to enable unconventional computing beyond the capabilities of von-Neumann computers. By merging brain-inspired architectures with concepts from quantum physics, we aim to implement versatile architectures for quantum computing, quantum communication and quantum simulation.
Group: Prof. Wolfram Pernice
Group: PD. Johannes Schemmel
Microelectronic circuits are developed, tested and applied. These microchips often contain extremely sensitive, low noise amplifiers for capturing sensor data and modules for further analog and digital signal processing. The crucial parts of such chips are designed completely manually. They are simulated on the analog level to achieve a maximal performance. The designs are fabricated in state-of-the-art CMOS technologies and are put into operation here at the group. A typical use case consists not only of designing the chip, but also includes building suitable control and data acquisition systems, the control and synchronisation of all components and the analysis of the measured data.
Recent developments include highly integrated circuits for positron emission tomography, readout electronics for DEPFET sensors for the future ILC detector, chips for detecting X-rays with hybrid pixel sensors, novel monolithic pixel sensors, development of front-end electronics for the CBM experiment at FAIR at the GSI, high-speed microscopy within the Viroquant project, detectors for synchrotron experiments at DESY, ESRF and the future XFEL, and circuit design techniques for generation of secret keys for cryptography.
In the group of Prof. Hamprecht, the topics of research include learning algorithms for image analysis, and its applications to the segmentation of biological images and beyond and tracking.
Prof. Rother's interests also lie in the field of computer vision and machine learning - ranging from deep learning and graphical models to smart data generation. A broad range of applications are investigated, such as image editing, image matching, scene understanding and bio-imaging.