European photonics collaboration to revolutionise brain-inspired computing
An international consortium including experts from the Zepler Institute for Photonics and Nanoelectronics is creating energy and size efficient plasmonic circuits for a new generation of neuromorphic computing.
The multi-million euro PlasmoniAC project will produce brain-inspired hardware that can learn from its inputs in ways not possible through contemporary computers.
The result could be an entirely new class of machines capable of being trained to recognise patterns using fewer inputs and far less energy, enabling a new type of learning capable mobile devices and sensors.
The three-year project, which is funded by the European Union's Horizon 2020 programme, brings together 10 industrial and academic partners from seven countries.
Dr Frederic Gardes, Optoelectronics Research Centre (ORC) Associate Professor, says: "I am delighted that world-leading research centres across Europe have been given the opportunity to come together to develop learning capable hardware. I believe the platform developed in PlasmoniAC will extend beyond its proposed application and enable researchers in a number of fields to have significant impact in a variety of CMOS-based platforms.
"We are targeting a powerful artificial photonic plasmonic neuron suite with up to three orders of magnitude higher computational efficiencies per neuron and between one and six orders of magnitude higher energy and footprint efficiencies, respectively, compared to the top state-of-the-art neuromorphic machines."
ORC researchers at the University of Southampton will provide photonic building blocks for the project based on a novel material developed in the Zepler Institute cleanrooms.
The computing industry is rapidly needing to move on from current CMOS electronics and von-Neumann architectures which cannot keep the pace with the requirements of new computational power metrics.
A new computing paradigm based on non-von-Neumann layouts, exploiting the working principles of the human brain, has already begun to unfold, leading to the development of large neuromorphic machines that already exceed the overall efficiency of classical platforms.
PlasmoniAC will build a future-proof solution based on plasmonics, a natural platform for synergising photonic-level bandwidths with electronic-level sizes within an ultra-high energy efficiency envelope.
The consortium includes four distinguished universities, Aristotle University of Thessaloniki in Greece, the University of Southampton, the Swiss Federal Institute of Technology in Zurich, and the University of Burgundy in France.
Industrial partners include the French National Centre for Scientific Research, the Interuniversity Microelectronics Centre in Belgium, IBM Research in Zurich, AMO in Germany, Israeli-American multinational Mellanox Technologies Ltd and VPIphotonics in Germany.
The new plasmonic circuit technology will place the European photonics industry in a world-leading position in the global neuromorphic and deep learning market.