A microchip is pictured on a woman's finger during a presentation of the German Bundesdruckerei
A microchip is pictured on a woman's finger during a presentation of the German Bundesdruckerei (German Federal Print Office) and the Fraunhofer-Institut for Reliability and Micro Integration (IZM) in Berlin July 11, 2007. The chip, which is less than 10 micrometres thick, will in future be used in paper-based security documents like passports. Reuters/Arnd Wiegmann

A team of Chinese researchers has created a type of biologically-inspired artificial neural network (ANN) called spiking neural network (SNN). The SNN comes in the form of a chip called “Darwin,” which mimics the principles of biological brain to perform faster data processing.

The researchers from Zhejiang University and Hangzhou Dianzi University in China believe that the Darwin neural processing unit (NPU) could lead to development in Internet of Things and other artificial intelligence systems. Darwin NPU processes information based on discrete-time spikes.

According to the researchers, the Darwin chip is an SNN-based neuromorphic hardware co-processor. It has been fabricated by standard CMOS technology. The research team says that SNN is more biologically-realistic than the traditional ANNs. In addition, it can potentially achieve much better performance-power ratio.

The Darwin NPU is hardwired to provide hardware acceleration of intelligent algorithms. Adding a boon for low-power and resource-constrained small embedded devices, the Darwin NPU has been fabricated by 180nm standard CMOS process. The latter supports up to 2048 neurons, in addition to 15 possible synaptic delays and more than 4 million synapses.

The best characteristic about Darwin NPU is that its SNN topology can be reconfigured. That is, it can be modified to include different configurations of synapses and neurons. It can be configured in multiple ways by its user to include different functionalities.

The successful development of Darwin chips shows that it is very much feasible to execute SNN in resource-constrained embedded systems. The Times of India reports that “since it uses spikes for information processing and transmission, similar to biological neural networks, it may be suitable for analysis and processing of biological spiking neural signals, and building brain-computer interface systems by interfacing with animal or human brains.”

Contact the writer at feedback@ibtimes.com.au, or let us know what you think below.

RELATED STORIES