In this talk I will present an ultrafast spatio-temporal optical computing system that exploits the dispersive properties of a step-index few-mode fiber to perform high-speed coincidence detection and address nonlinear computational tasks. Linear mode mixing within the fiber provides short-term memory for time-encoded signals, while output features extracted from the spatial intensity distribution are used for classification. Building on previous demonstrations of high-speed linear classification, I investigate the impact of a gain-saturation nonlinearity at the post-processing stage for solving nonlinear benchmark tasks, including delayed XOR and parity XOR. The insights gained from these results will guide the experimental implementation of a nonlinear optical computing system, in which the gain saturation of a semiconductor optical amplifier is exploited prior to photodetection to enhance performance on nonlinear classification tasks.
Contact details:
Apostolos Argyris Contact form