Fig. 4: Scaling exponents of the proposed method.
From: Scaling advantage of chaotic amplitude control for high-performance combinatorial optimization
Scaling exponents \(\frac{\gamma }{\,{{\mbox{log}}}\,(10)}\) of the 50th (a) and 80th (b) percentiles of the time to solution distribution based on the hypotheses of scaling in eγN obtained by fitting data of the number of matrix-vector multiplication (MVM) to solution vs. problem size N shown in Fig. 2 of the proposed chaotic amplitude control dynamics and the scaling exponents reported in ref. 20. Colored boxes show the 90% confidence interval in the scaling exponents. CAC: chaotic amplitude control; SA: simulated annealing; PT: parallel tempering; DA: digital annealer; PTDA: parallel tempering on DA. Exponents of SA, PT, DA, and PTDA are taken from ref. 20. CAC (100 × 100) and CAC (fully parallel) are estimated using time (see Fig. 3) and MVM (see Fig. 2) to solution, respectively.
