Photonic Extreme Learning Machines Using Hexagonal Programmable Waveguide Meshes
José Roberto Rausell‐Campo, Daniel Pérez, José Capmany Francoy · 2024
We propose employing a hexagonal mesh as a Programmable Photonic Extreme Learning Machine. Input encoding, random matrix multiplication and nonlinearity are performed on-chip while trained digitally. Experimental validation is obtained on two classification tasks, the Iris and Banknote dataset, with >95% and >91% accuracy.