Integrated laser graded neuron enabling high-speed reservoir computing without a feedback loop

Photonic neuromorphic computing is a promising paradigm for next-generation high-speed and energy-efficient computing. Photonic neurons, as the key components to empower computational expressiveness, pose a significant bottleneck in speed and efficiency. Currently, the widely studied photonic spiking neurons suffer from limitations in signal processing speed due to their refractory period, and most of them are optical-injected, requiring additional laser sources and modulators. Here, we experimentally demonstrate a bio-inspired integrated photonic graded neuron based on an electrical-injected quantum-dot laser. With a different signal injection method, the laser can exhibit the graded response and avoid the pulse response rate limitation caused by the refractory period, thereby achieving high processing speed. The laser neuron does not require external laser pumps and modulators, leading to a simple and low-energy-consuming system. Leveraging the powerful memory effect and nonlinear dynamics of the laser graded neuron, we build a reservoir computing (RC) system that eliminates the need for any time-delayed feedback loop. Our RC system shows excellent performance in MIT-BIH arrhythmia detection and MNIST classification tasks with accuracies of 98.4% and 92.3%, respectively, and a Mackey-Glass prediction task with a normalized root-mean-square error (NRMSE) of 0.108 for a 30-step prediction. This work opens a route for high-speed photonic neuromorphic computing.