Quantamagazine covers our work on learning on analog neuromorphic hardware

Allison Whitten wrote a nice article for Quantamagazine about our recent collaboration with Heidelberg on the BrainScaleS-2 neuromorphic chip. Analog hardware faces similar obstacles as the brain when learning: Neuronal heterogeneity means that you won’t find two neurons that are the same. This creates special demands on the learning systems and algorithms that instill functions into such substrates. Our work sketches a path to overcome this issue with bio-inspired learning rules. Read Allison’s article here: https://www.quantamagazine.org/ai-overcomes-stumbling-block-on-brain-inspired-hardware-20220217/

The BrainScaleS-2 neuromorphic chip (University of Heidelberg) in action after being trained with in-the-loop training using surrogate gradients.