Science Advances

This PDF file includes:

• Section S1. D-Wave embeddings and J_c optimization
• Section S2. CIM data and post-selection
• Section S3. C-SDE simulations of CIM
• Section S4. Optimal–annealing time analysis
• Section S5. Performance of parallel tempering
• Fig. S1. D-Wave success probability for SK problems and MAX-CUT problems of edge density 0.5 as a function of problem size N and embedding parameter J_c.
• Fig. S2. MAX-CUT on graphs with an edge density of 0.5.
• Fig. S3. Properties of heuristic embeddings for fixed-degree graphs.
• Fig. S4. Choice of optimal coupling for sparse graphs using the heuristic embedding.
• Fig. S5. Data filtering and post-selection in NTT CIM.
• Fig. S6. Comparison of Stanford and NTT CIM performance for SK and dense MAX-CUT problems.
• Fig. S7. Abstract schematic of measurement-feedback CIM.
• Fig. S8. Simulated CIM success probability as a function of F_max for the SK, dense MAXCUT, and cubic MAX-CUT problems in this paper.
• Fig. S9. Comparison of c-SDE simulations with experimental CIM data.
• Fig. S10. Simulated CIM success probability and time to solution (in round trips) for SK and MAX-CUT problems.
• Fig. S11. Time-to-solution analysis for D-Wave at optimal annealing time.
• Fig. S12. CIM time to solution compared against the parallel tempering algorithm implemented in the Unified Framework for Optimization.
• Table S1. Seven steps in a single round trip for the measurement-feedback CIM and the appropriate truncated Wigner description.
• Table S2. Problem-dependent constants α and β used in the relation N₀ = α + β log₁₀(T/μs) for the success probability exponential P = e^{−(N/N₀)2.}
• ^{References (61–64)}

^{Download PDF}