High-Speed Reservoir Computing With Integrated Laser Graded Artificial Neurons
1 min read
Summary
Research by scientists at Swinburne University of Technology in Australia and Shanghai University in China has enabled the creation of photonic neurons that operate much like the natural neurons in the human brain.
The breakthrough in so-called neuromorphic computing comes after previous models were limited by a binary response system and a refractory period that meant the neurons were only capable of an on or off function.
The new model uses quantum dot lasers integrated on a single die that enable an analog, or graduated, response without a refractory period.
The team tested the new device using reservoir computing in tasks including arrhythmia detection (98% accuracy) and handwriting classification (92% accuracy).
The development could usher in a new era of ultra-low power, fast integrated brains for computing.
