Fig. 2: Quantum annealer inverse temperature estimation.
From: Conditioned quantum-assisted deep generative surrogate for particle-calorimeter interactions
a Mean number of iterations to meet the reduced standard error threshold in Eq. (1) using different iterative methods. Method 1 uses the KL divergence as in Eq. (23). Method 2 and 3 use Eq. (24), in addition Method 2 adapts the δ parameter after each iteration such that λ ≈ 0 (see Eq. (25)). The dashed purple line corresponds to the ratio between final minus initial effective β and the number of iterations, such that higher values implies faster convergence. The bars correspond to the standard deviation. b Estimated inverse temperature vs iterations using method 2, adaptive. c RBM histogram using classically generated and QA-generated samples after estimated temperature convergence.
