Photonic Neuromorphic Pattern Recognition with a Spiking DFB‐SA Laser Subject to Incoherent Optical Injection

Abstract

Photonic neuromorphic computing is a competitive paradigm to overcome the bottleneck of von Neumann architectures. Incoherent and coherent synaptic networks are two popular schemes realizing photonic weighting functions. Previous works have proved the distributed feedback (DFB) laser with an intracavity saturable absorber (DFB‐SA) can behavior like a spiking neuron. However, the compatibility with the incoherent synaptic architecture has not yet been demonstrated. Here the neuron‐like dynamics of a DFB‐SA laser subject to single‐wavelength and multiple‐wavelengths incoherent optical injections are experimentally demonstrated. The results show that, for the DFB‐SA laser subject to single‐wavelength incoherent injection, the neuron‐like dynamics including threshold, temporal integration, and refractory period are achieved. Besides, the range of injection wavelength that leads to a successful neuron‐like response is identified. For the DFB‐SA laser with four‐wavelength incoherent optical injection, the neuron‐like dynamics can also be achieved. In addition, the effect of wavelength interval is also considered. The logic XOR operation and Iris recognition tasks are successfully implemented. Furthermore, the feasibility of a cascaded system for the DFB‐SA lasers with four‐wavelengths incoherent optical injection is demonstrated. This work provides a feasible scheme for the system integration of photonic spiking neurons and incoherent synaptic networks.