Belgian researchers have created a chip for optical information processing using techniques inspired by how the human brain works.
During an experiment described in a recent issue of the Nature Communications journal, the team from the University of Ghent created a small neural network consisting of 16 nodes embedded in a silicon photonics chip.
The main advantage of the chip is the fact that it combines the same technology used in common computer chips but relies on light instead of electricity as the information carrier, opening up possibilities for extremely fast computing devices with substantially lower power consumption.
The researches demonstrated such a chip could perform a whole variety of tasks like arbitrary calculations with memory on a bit stream or header recognition in telecommunications. Header is operation relevant in telecommunication networks indicating where data needs to be sent.
Further simulations have shown that the same chip can perform a limited form of speech recognition, by recognising individual spoken digits.
So was, neural networks have been employed to solve pattern recognition problems like speech recognition or image recognition. However, most of the existing neural network systems have been implemented as part of software on a traditional computer.