Photonic extreme learning machine by free-space optical propagation

Photonic brain-inspired platforms are emerging as novel analog computing devices, enabling fast and energy-efficient operations for machine learning. These artificial neural networks generally require tailored optical elements, such integrated photonic circuits, engineered diffractive layers, nanophotonic materials, or time-delay schemes, which challenging to train stabilize. Here, we present a neuromorphic scheme, i.e., the extreme learning machine, can be implemented simply by using an encoder coherent wave propagation in free space. We realize concept through spatial light modulation of laser beam, with far field acting feature mapping experimentally demonstrate from data on various classification regression tasks, achieving accuracies comparable digital kernel machines deep networks. Our findings point out device that is easy train, energetically efficient, scalable, fabrication-constraint free. The scheme generalized plethora systems, opening route real-time processing data.