An unsupervised deep learning algorithm for single-site reconstruction

We have developed a novel ML-inspired algorithm for high-fidelity reconstruction in extreme regimes.

We propose a novel technique based on a convolutional autoencoder which enables us to reconstruct the lattice occupation with high fidelity even though our lattice spacing is more than two times smaller than our imaging resolution (383.5nm lattice spacing vs. 850nm Rayleigh resolution).