Feedback-free photonic reservoir computer based on a two-element laser array

Photonic reservoir computing (RC) is powerful but limited by the integration and stability challenges of time-delayed feedback loops. To address this limitation, we propose a feedback-free quasi-convolutional encoded (QC) reservoir computer (QRC) based on a two-element semiconductor laser array. In this system, the original masked input and the quasi-convolutionally encoded input are injected into two laterally coupled lasers, and the transient laser responses are used as the reservoir states. Numerical simulations across four waveguide structures and three benchmark tasks show that the proposed system provides broad parameter tolerance and stable operation without a physical feedback loop. The accuracy advantage over a conventional time-delayed reservoir computer is task- and parameter-dependent, whereas the most robust benefit is the enlarged stable operating window under variations of injection rate, detuning, pump rate, and laser separation. Dynamical analyses indicate that the feedback-free laser-array system mitigates feedback-induced periodic and chaotic instabilities, enabling more reproducible transient-state generation. These results suggest that QC-assisted feedback-free laser-array RC is a promising approach for compact and integration-friendly neuromorphic photonic computing.