Fig. 2: Binary sparse coding image reconstructions of the original image shown in Fig. 1. Top left: The best (lowest energy) reconstruction found using D-Wave with standard forward annealing (with no parallel QA and an annealing time of 20 microseconds). | npj Unconventional Computing

Fig. 2: Binary sparse coding image reconstructions of the original image shown in Fig. 1. Top left: The best (lowest energy) reconstruction found using D-Wave with standard forward annealing (with no parallel QA and an annealing time of 20 microseconds).

From: Comparing quantum annealing and spiking neuromorphic computing for sampling binary sparse coding QUBO problems

Fig. 2

Bottom left: reverse quantum annealing chain of Monte Carlo-like iterations (QEMC). Top right: Single execution of neuromorphic computing on Loihi 2. Bottom right: Iterated warm-starting neuromorphic computing with Loihi 2. Combined, the bottom row of plots shows that the two iterative approaches yield better solution quality than that non iterative improvement methods shown in the top row. All experimentally computed figure reconstructions are the best mean energy and best mean sparsity across the 16 QUBO models (for the best parameter combination found for each device or technique).

Back to article page