Phase and Amplitude Encoding for all Optical Kernel Methods with a Photonic Lantern

Our computational needs have reached unprecedented levels, accompanied by a significant increase in power consumption. This has spurred growing interest in exploring novel hardware approaches including optical methods to address these challenges [1]. One promising strategy involves kernel methods, which map the input data into high-dimensional spaces, enabling linear algorithms to solve nonlinear problems efficiently [2]. This paper presents two main contributions: i) the use of a novel fiber component - a photonic lantern - as the physical kernel that performs the optical mapping and ii) a comparison between two different data encoding techniques - phase and amplitude of a laser beam on the accuracy of the classification tasks.