Researchers have developed a laser-based artificial neuron that fully mimics the function, dynamics and information processing of biologically graded neurons. The new laser-graded neurons can process signals at 10 GBaud, a billion times faster than their biological counterparts, and could lead to breakthroughs in areas such as artificial intelligence and other types of advanced computing.
The body contains various types of nerve cells, including graded neurons, which encode information through continuous changes in membrane potential, allowing for subtle and precise signal processing. In contrast, biological spiking neurons use all-or-no action potentials to transmit information, creating a more binary form of communication.
“Our laser graded neurons overcome the speed limitations of current photon versions of spiking neurons and have the potential to run even faster,” said Chaoran Huang, leader of the research group at the Chinese University of Hong Kong. “By taking advantage of their neuron-like non-nature With linear dynamics and fast processing, we built a reservoir computing system that demonstrates superior performance in artificial intelligence tasks such as pattern recognition and sequence prediction.”
exist OpticalIn Optica Publishing Group’s high-impact research journal, the researchers report that their chip-based quantum dot laser-graded neurons can achieve signal processing speeds of 10 GBaud. They used this speed to process data from 100 million heartbeats or 34.7 million images of handwritten digits in one second.
“Our technology can accelerate AI decision-making in time-critical applications while maintaining high accuracy,” Huang said. “We hope that integrating our technology into edge computing devices (processing data close to the data source) will facilitate faster , smarter artificial intelligence systems to better serve real-world applications in the future while reducing energy consumption.”
Faster laser neurons
Laser-based artificial neurons that respond to input signals in a way that mimics the behavior of biological neurons are being explored as a way to significantly enhance computing due to their ultra-fast data processing speeds and low energy consumption. However, most neurons developed to date are photon-spike neurons. These artificial neurons have limited response speeds, may lose information, and require additional laser sources and modulators.
The speed limit of photon-spike neurons comes from the fact that they typically work by injecting input pulses into the gain portion of the laser. This causes a delay, limiting how quickly the neuron can respond. For laser-graded neurons, the researchers took a different approach and avoided this delay by injecting radiofrequency signals into the saturable absorbing portion of the quantum dot laser. They also designed high-speed RF pads for saturable absorption sections to produce faster, simpler and more energy-efficient systems.
“With strong memory effects and excellent information processing capabilities, a single laser-graded neuron can behave like a small neural network,” Huang said. “Therefore, even a single laser-graded neuron without additional complex connections can perform high-performance Perform machine learning tasks.”
High-speed reservoir calculation
To further demonstrate the capabilities of the laser-graded neurons, the researchers used it to create a reservoir computing system. This computing method uses a special type of network called a “reservoir” to process time-related data, such as that used for speech recognition and weather forecasting. The neuron-like nonlinear dynamics and fast processing speed of laser graded neurons make them ideal for supporting high-speed reservoir calculations.
In tests, the resulting reservoir computing system demonstrated excellent pattern recognition and sequence prediction in a variety of artificial intelligence applications, especially long-term prediction, with fast processing speed. For example, it processes 100 million heartbeats per second and detects arrhythmia patterns with an average accuracy of 98.4%.
“In this work, we used a single laser-graded neuron, but we believe that cascading multiple laser-graded neurons will further unlock their potential, just as the brain has billions of neurons working together in a network ,” Huang said. “We are working to increase the processing speed of laser-grading neurons while developing deep computational architectures containing cascaded laser-grading neurons.”
