On-chip neural network uses silicon photonics
Researchers from Ghent University have created an on-chip neural network using silicon photonics and say the design was inspired by the way in which the brain works.
While neural networks have been employed to solve such problems as speech and image recognition, the approach has generally been implemented in software on a traditional computer. The Ghent researchers implemented a 16 node neural network directly in hardware, using silicon photonics. This approach, they said, has the potential for extremely high computation rates and low power consumption.
In their experiments, reported in Nature Communications, the researchers have also shown the chip can be used for a range of tasks, including arbitrary calculations, header recognition and a limited form of speech recognition.
