EBRAINS offers brain-inspired tools and services to understand and leverage the computational capabilities of spiking neural networks. Unlike standard deep neural networks, which require considerable amounts of energy and data, spiking neural networks are key to understanding the human brain's ability to learn continuously and implement higher cognitive functions, while consuming only 20W.
EBRAINS neuromorphic computing systems provide larger spiking neural network simulations and emulations through specialised, high-performance, low-power hardware systems. BrainScaleS emulates spiking neural networks up to 1,000 times faster than real time, with fixed analogue neurons and programmable plasticity. SpiNNaker digitally simulates large spiking neural networks in biological real time. While BrainScaleS is especially suitable for research into learning and synaptic plasticity, and for network training, SpiNNaker is particularly relevant for robotic applications, thanks to its programmatic versatility and an existing interface with the Neurorobotics Platform (NRP).
The NRP is the integrative simulation service of EBRAINS that provides brain simulation with the dimensions of embodiment and closed-loop interactions with a virtual environment. This allows for multiple learning strategies to be implemented, from reinforcement learning to unsupervised learning with spike-timing-dependent plasticity, making the NRP an ideal platform for virtual prototyping and developing applications based on neuromorphic computing.