BrainScaleS 2
BrainScaleS 2 is a neuromorphic computing system designed to emulate large-scale spiking neural networks (SNNs) in real-time, accelerating research into brain-inspired computation. Current research focuses on improving learning algorithms for SNNs, particularly developing efficient event-based backpropagation methods and exploring the impact of incorporating synaptic delays into network models, often leveraging the system's analog hardware capabilities for faster and more energy-efficient training. This platform facilitates investigations into SNN dynamics and learning, offering valuable insights into biological neural computation and potentially paving the way for more energy-efficient artificial intelligence.
Papers
Demonstrating the Advantages of Analog Wafer-Scale Neuromorphic Hardware
Hartmut Schmidt, Andreas Grübl, José Montes, Eric Müller, Sebastian Schmitt, Johannes Schemmel
Reproduction of AdEx dynamics on neuromorphic hardware through data embedding and simulation-based inference
Jakob Huhle, Jakob Kaiser, Eric Müller, Johannes Schemmel