Researchers from Gent University, Belgium, have demonstrated optical information processing using a light-based chip inspired by the way our brain works.
Scientists have looked to neural networks before, mostly to develop bio-inspired techniques implemented in software on a traditional computer, for example, to solve pattern recognition problems like speech recognition or image recognition. However, the Ghent researchers integrated a small neural network directly into hardware, using a silicon photonics chip, which uses light rather than electricity as the information carrier. “This silicon photonics chip is inspired by the way our brain works, which opens the way for information processing at high speeds (up to several hundreds of gigabits per second) and low power consumption,” says Professor Peter Bienstman of Ghent’s Photonics Research Group.
As the Ghent experiments have shown, the chip could be used for a variety of tasks, from arbitrary calculations with memory on a bit stream or header recognition to recognising individual spoken digits. So far, Bienstman confirms, “This prototype can recognise bit sequences in a bit stream, as well as perform arbitrary boolean operations on a bit stream.”
Next, Bienstman wants to scale it up the current silicon photonics chip, which only has 16 neurons, to larger neural networks. “This should allow us to tackle more complex applications, such as basic building blocks for speech recognition, channel equalisation and error correction for telecom, etc.” he concludes.
Ghent’s research work is detailed in the paper “Experimental demonstration of reservoir computing on a silicon photonics chip,” published in Nature Communications.
The above story is based on materials by OPTIMAX.
Note: Materials may be edited for content and length. For further information, please contact the source cited above
See more at: http://www.novuslight.com/silicon-photonics-chip-inspired-by-brain_N2488.html#sthash.5dPuJQld.dpuf