[IEEE 2019 Conference on Lasers and Electro-Optics Europe & European Quantum Electronics Conference (CLEO/Europe-EQEC) - Munich, Germany (2019.6.23-2019.6.27)] 2019 Conference on Lasers and Electro-Optics Europe & European Quantum Electronics Conference (CLEO/Europe-EQEC) - Phase Noise Robustness of a Coherent Spatially Parallel Optical Reservoir

摘要： Reservoir computing is a machine learning technique that is particularly adapted to process time dependent information. It is also very well suitable to experimental implementations, and in particular photonic implementations. In the present work we propose strategies to improve the phase noise robustness of a spatially parallel reservoir computer based on a coherent photonic cavity without active stabilization. The system, described in [4], is currently under development in our laboratory. It is based on a linear Fabry-Pérot resonator, with neurons encoded as a grid of 9 by 9 focused spots on the input mirror plane, and a nonlinear readout operating at 0.5GHz. The coupling between neurons is realized by a Spatial Light Modulator (SLM) placed at the back plane of the cavity and is optimized for each task. We present numerical results obtained for the 4-level channel equalization task, and we find similar results for the NARMA10 task.