In a recent publication in Nature Communications researchers from Ghent University report on a new technique for optical information processing on chips using techniques inspired by the way our brain works.
Although neural networks have been used in the past to solve pattern recognition problems such as speech and image recognition, it was usually in software on a conventional computer.
Ghent University researchers have manufactured such a small neural network (16 nodes) in hardware, using a silicon photonics chip. This chip is made using the same technology as traditional computer chips but uses light instead of electricity as information carrier. This approach has many advantages, among others as potential for high speed and low power consumption.
The scientists have experimentally shown that the same chip can be used for a diverse range of tasks, such as calculations with memory on a bit stream and so-called ‘header recognition’. In addition, simulations have shown that the same chip can perform limited speech recognition recognition of indivivuele spoken digits.
This study was done in a collaboration between members of the Photonics Research Group of the Department of Information Technology (among them Peter Bienstman) and the Reservoir Lab of the Department of Electronics and Information Systems. It was funded by the European Research Council (ERC) Starting Grant from the Naresco and by the Belgian IAP program via the Photonics@BE network.