Few-Shot Neuromorphic Vision in a Nonlinear Photonic Network Laser
This work introduces a photonic computing system inspired by biological vision, using spatially-competing lasing modes in a laser network to detect and classify features from images. By combining excitatory and inhibitory nonlinear dynamics, the system achieves strong performance in few-shot learning and low-data scenarios, outperforming software neural networks on tasks like digit recognition and cancer diagnosis.