Skip to main content
. 2015 Jan 16;55(3):529–537. doi: 10.1021/ci500749q

Figure 4.

Figure 4

Comparison of property-optimizing ACSESS outcomes to the simple genetic algorithm (SGA) optimization for generating multiple solutions with the largest NKp fitness value. SGA iteratively samples the fittest solutions and improves upon them. (A) indicates that SGA finds the fittest solutions (global maximum in the constructed NKp model) after 30 iterations. (B) indicates that property-optimizing ACSESS runs escapes local optima more effectively than do the SGA runs, i.e. the success rate of finding largest fitness solution of property-optimizing ACSESS is 100% compared to 60% for SGA (from 10 runs of each method). (C) indicates that the property-optimizing ACSESS runs also performs better than SGA to sample diverse fittest solutions as it samples a larger number of fittest solutions compared to SGA. (D) indicates that, on average, the fitness of solutions generated by property-optimizing ACSESS is larger than that of SGA.