Intelligent Control of Mode-Locked Fiber Laser via Residual Neural Network

Nonlinear polarization rotation (NPR) mode-locking technology has emerged as a preferred approach for many researchers in implementing passive mode-locked fiber lasers (PMLFLs) This preference stems from its simplicity, flexibility, high performance and the rich nonlinear dynamics it introduces into the laser cavity. With flexible polarization control, NPR-based mode-locked lasers can generate diverse pulse states, such as Q-switching (QS) pulses, Q-switched mode-locking (QML) pulses, fundamental mode-locking (FML) pulses and harmonic mode-locking (HML) pulses. However, achieving the desired pulse state immediately after power-on and maintaining it over extended periods remains challenging for NPR-based PMLFLs. Recently, intelligent mode-locked lasers have garnered significant attention. Researchers have leveraged intelligent algorithms to the mode-locked laser to enable precise and adaptive regulation of mode-locked states [1], [2]. In this study, we propose a residual neural network specialized for spectral accurate classification and apply it to NPR-based PMLFLs. This approach enables reproduction of desired mode-locked state spectra. Our findings aim to address the challenges associated with NPR mode-locked lasers, thereby expanding their potential applications.