ACA: towards multi-scale natural-density Neuromorphic Computing
The pilot project ACA (Advanced Computing Architectures) combined expertise at the Jülich Research Centre, the RWTH Aachen University, Heidelberg University and the University of Manchester to drive the research on Neuromorphic Computing forward. The project, which was funded by the Helmholtz Initiative and Networking Fund and the Jülich Research Centre, ran from November 2018 to October 2022. The research initiated inside the project is continued in the frame of the Jülich Neuromorphic Computing Alliance (JUNCA).
In tasks such as object recognition in natural environments, unsupervised learning from few examples and with little reward, or prediction of future actions of other individuals, the brains of humans and other animals outperform traditional computers with respect to computational capacity, robustness against noise and malfunction, processing speed, as well as material cost, size, and energy efficiency. Neuromorphic Computing aims at developing advanced computing architectures with brain-like performance characteristics by exploiting the principles employed by nature. It refers to a new form of information processing based on technologies and algorithms mimicking the structure and dynamics of biological neuronal systems.
Neuromorphic Computing addresses two different application areas: Cognitive computing and Neuroscience simulation. Cognitive computing refers to the field of machine learning, deep learning, and artificial neural networks, i.e. algorithms and technologies for general purpose applications such as early disease diagnostics or robotics. Neuroscience simulation is used as a tool to study the structure, dynamics and function of biological neuronal systems, and thereby to uncover the principles underlying their superior performance.
This project is targeting the Neuroscience simulation application area. It is a pilot project preparing a long-term Neuromorphic Computing research initiative. Its main goal is the specification of a future Neuromorphic Computing architecture, including the definition of requirements and target performances, the development of workflows for a systematic validation and benchmarking of neuromorphic architectures, and the development of efficient Neuromorphic Computing concepts, e.g. for the instantiation of and communication within complex neuronal networks or for implementations of single-neuron and synapse dynamics.
The project is organized into four work packages (WPs) according to the central problem areas:
The project "Advanced Computing Architectures (ACA): towards multi-scale natural-density Neuromorphic Computing" wa a pilot project funded by the Helmholtz Initiative and Networking Fund (project no. SO-092), and the Jülich Research Centre (Nov 2018 - Oct 2022).