Image classification with a single spiking microlaser

This work presents a comprehensive study on image classification using a single spiking microlaser within a neuromorphic photonic framework. The research leverages a vertical-cavity surface-emitting laser (VCSEL) with an integrated saturable absorber, which exhibits excitable dynamics analogous to biological neurons. Operating in the excitable regime, the microlaser emits spikes in response to perturbations, enabling its use as a standalone nonlinear node in a reservoir computing (RC) architecture. The system is analysed both from numerical and experimental perspectives. The sparse spiking output coming from the system is decoded into binary vectors via time-binning, and the readout matrix is trained using a simple Moore-Penrose pseudoinverse optimization, seeking to minimize, and ultimately remove, dependence on an external computer. An experimental characterization confirms the excitable behaviour of the microlaser. This work demonstrates the viability of spiking microlasers as ultra-fast, energy-efficient computational units for neuromorphic photonics.