Fig. 2: Experimental run and data processing.
From: Integrated photonic programmable random matrix generator with minimal active components
a Sequences of testing random pulsed trains \({{\bf{x}}}^{(p)}\in {{\mathbb{R}}}^{100}\), for p ∈ {1, …, 20}. b The latter are demultiplexed into signals \({\hat{{\bf{x}}}}^{(p)}:= {{\mathbb{R}}}^{20\times 5}\), which are programmed in the N × 1 switch and injected into the PIC. This produces the randomized demultiplexed signals \({\hat{{\bf{z}}}}^{(p)}\). c In this process, two ensembles of random phases are loaded into the phase shifters through the SMU, which correspondingly powers the metal heaters. Such ensembles are shown as histograms and scatter plots, highlighting the normal (left), uniform (right), and Rayleigh (center) distribution profiles. d The output processed signals from the PIC are multiplexed back to the vectors \({{\bf{z}}}^{(p)}\in {{\mathbb{R}}}^{100}\). e–g These outputs are then post-processed to extract the statistical information for normally and uniformly randomized phase distributions using the entropy S[∣x∣] and autocorrelation difference ΔℓX[∣x∣] criteria. For reference purposes, the typical entropy values for normal and uniform distributions are highlighted in yellow and blue, respectively. Likewise, the typical autocorrelation values for the previous distributions are highlighted in red.
