NICOP - BRAIN-inspired networks of ultrafast LASER neurons
Abstract
This research programme will use semiconductor lasers, the very same devices used to transmit high-speed internet data traffic over fibre-optic telecommunication systems, to develop networks of photonic neurons inspired by the brain’s powerful computational capabilities and able to perform complex information processing tasks at ultrafast speeds. Biological neurons are cells that compute information in the brain by firing spikes when stimulated. Quite remarkably, semiconductor lasers can also produce a rich variety of spiking responses and other complex dynamic behaviours similar to those observed in biological neurons but up to 1 billion times faster. Moreover, semiconductor lasers are also compact devices, easy to integrate into photonic chips and networks, and fully compatible with current optical telecommunication technologies. These interesting features make them ideal candidates for their future use as photonic neuronal models and to become the building blocks of highly interconnected networks of photonic neurons with neuro-inspired connectivity. Hence, photonic neuronal models based on semiconductor lasers offer high potentials for novel brain-inspired computing modules using light signals to perform information processing tasks in the same way neurons operate in the brain but at ultrafast speeds. This ambitious research programme in neuromorphic photonics will focus on three transformative objectives, namely (i) the investigation of versatile and ultrafast photonic neuronal models using different types of semiconductor laser structures; (ii) their scaling into network architectures with neuro-inspired connectivity and (iii) the delivery of proof-of-concept demonstration of brain-like photonic information processing at ultrafast speeds up to sub-nanosecond regimes (where a nanosecond equals a billionth of a second). These breakthroughs will open exciting routes for novel neuromorphic photonic systems transcending classical digital computing systems, going beyond the state-of-the-art, and allowing to process information as neurons do in our brain but up to 9 orders of magnitude faster than the millisecond timescales of biological neurons. These will find use in future computing systems, communication networks and applications requiring analysis/classification of large amounts of data (e.g. military/security, healthcare, neuroscience and online social networks among others). This inter-disciplinary and timely research project brings together the hitherto disparate fields of semiconductor lasers, nonlinear dynamics, neuromorphic engineering and photonics. Following this interdisciplinary research approach we arrive at the long-term vision of networks of photonic neurons as a future key enabling technology for ultrafast, energy efficient information processing inspired by the brain’s powerful computational capabilities and with ample impact across disparate fields (e.g. optical networks, big data analysis, ultrafast temporal pattern recognition and decision making). This is a very innovative aspect and there is much to be gained from this cutting-edge programme. Finally, the applicants’ track record and expertise in neuromorphic photonics, semiconductor lasers, laser dynamics and optoelectronics added to the capabilities at the Hosting Institution (Institute of Photonics at the University of Strathclyde) will create a unique foundation for a wide research vision in neuromorphic photonics for novel paradigms in ultrafast brain-inspired computing going beyond classical digital information processing systems.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 20, 2017
- Source ID
- N629091812027
Entities
People
- Antonio Hurtado
Organizations
- Office of Naval Research
- United States Navy
- University of Strathclyde