Dynamical learning of a photonics quantum-state engineering process

Experimental engineering of high-dimensional quantum states is a crucial task for several information protocols. However, high degree precision in the characterization noisy experimental apparatus required to apply existing quantum-state This often lacking practical scenarios, affecting quality engineered states. We implement, experimentally, an automated adaptive optimization protocol engineer photonic orbital angular momentum (OAM) The protocol, given target output state, performs online estimation currently produced states, relying on measurement statistics, and determines how tune parameters optimize state generation. To achieve this, algorithm does not need be imbued with description generation itself. Rather, it operates fully black-box scenario, making scheme applicable wide variety circumstances. handles controlled by are rotation angles series waveplates can used probabilistically generate arbitrary four-dimensional OAM showcase our different both classical regimes prove its robustness external perturbations control parameters. approach represents powerful tool optimizations tasks protocols technologies.