Adaptive transmission matrix control for noise-robust optical information processing

Maintaining stable, high‑fidelity light transmission through dynamic scattering media is a key challenge for next‑generation optical communication and information processing. We present a hybrid feedback‑driven framework that continuously engineers the transmission pipeline and compensates for noise and decorrelation in real time. By integrating live intensity measurements into a neural‑network optimizer, our system adaptively updates the phase pattern on a Spatial Light Modulator without interrupting operation. In experiments, this approach sustains output correlations above 90 % for over an hour in an unstabilized diffusive environment - outperforming traditional offline methods. This self‑tuning platform lays a robust foundation for optical communications and enables applications in computing, high‑resolution imaging, and secure encryption.