Fig. 3
From: A continuous-time MaxSAT solver with high analog performance
Algorithm statistics over random Max 3-SAT problems. Distribution of differences between the real global minimum Emin obtained with the exact algorithm MaxSatz and a the smallest energy found by the algorithm \(\overline E\), b the predicted minimum value \(E_{{\mathrm{min}}}^{{\mathrm{pred}}}\), and c the final decision of the algorithm \(E_{{\mathrm{min}}}^{{\mathrm{dec}}}\) shown for problems with different N and α values (see legends). d The percentage of instances indicated by color (see color bar) for different values of the error \(E_{{\mathrm{min}}}^{{\mathrm{pred}}} - E_{{\mathrm{min}}}\) and hardness η. Most instances are in the \(E_{{\mathrm{min}}}^{{\mathrm{pred}}} - E_{{\mathrm{min}}} = 0\) row indicating correct prediction. Large errors occur mainly at smaller η values, and are dominantly negative
