Neuromorphic Computing Platform
The Neuromorphic Computing Platform will build on capabilities developed in the European FACETS and BrainScaleS projects and in the UK SpiNNaker Project. The current version of the Neuromorphic Physical Model (NM-PM) incorporates 50*106 plastic synapses and 200,000 biologically realistic neuron models on a single 8-inch silicon wafer in 180nm process technology. The system does not execute pre-programmed code but evolves according to the physical properties of the electronic devices. FACETS has also pioneered a network description language (PyNN) that provides platform independent access to software simulators and neuromorphic systems and will be used throughout HBP. BrainScales – a follow-up project – is pioneering the use of the technology to replicate behaviour and learning over periods of up to one year while simultaneously emulating the millisecond-scale dynamics of the system.
The Neuromorphic many-core system (NM-MC) will use the approach pioneered by the UK SpiNNaker group. The group has a strong grounding in the ARM architecture, which offers an excellent basis for scalable digital many-core systems operating at real time with low power, has recently completed the integration of a SpiNNaker chip into an operational system and is now running experiments. Each chip has eighteen cores and a shared local 128M byte RAM, and allows real-time simulation of networks implementing complex, non-linear neuron models. A single chip can simulate 16,000 neurons with eight million plastic synapses running in real time with an energy budget of 1W.
The operational phase of the HBP will see rapid scaling-up of the project's two Neurmorphic Computing Systems (the PM and the MC systems) and their supporting ecosystem (development tools, data analysis tools etc.). Development of the tools and technologies required to achieve this goal will begin during the ramp-up phase. By the end of this phase, HBP will have already developed software models (a simulation of the working system running on software) of the second version of the physical model (NM-PM2) and the many-core (NM-MC2) systems. These systems will thus be ready to enter manufacture at the start of the operational phase.
In parallel with this work, HBP will conduct exploratory studies of novel technologies necessary for the second and subsequent versions of the platforms, including high-density packaging technologies, and novel techniques of Computer Aided Design (CAD).
What People are Saying
Neuromorphic Computing is a huge opportunity for European science and industry. It could create a completely new industry based on those computing architectures [human brain inspired architectures] that are so energy efficient, fault tolerant and that do not use predefined software.
We've established collaborations with a number of groups around Europe and in fact the wider world. And the HBP has created a unique opportunity to extend these collaborations within Europe and to apply the SpiNNaker digital Neuromorphic computing technology to a wide range of interesting problems in the areas of computational neuroscience.
Within the open call part of the HBP, we're inviting new partners to join us to find suitable applications to explore the capabilities of the many core Neuromorphic technology that we've developed.
We're looking forward to receiving very original ideas from different fields that will make use of these novel computing systems - very much in the spirit of the way people used the first programmable computer built by John von Neumann in the Princeton Institute of Advanced study in the 40s of the last century.
The overall goal of the Neuromorphic Computing Subproject is to build and operate and use radically different computing architectures – Neuromorphic architecture – which is modeled after the architecture we find in the human brain and which we're going to explore in the human brain project.
Neuromorphic computers are systems radically different from traditional information processing devices in the sense that they really copy the structure, the function of biological brains.
The brain has amazing computing capabilities, it runs at very low power, it doesn't need any software, it's fault tolerant. It's pretty obvious that bringing these features into synthetic systems must have a rather severe effect on the way we build computers in the future.