Geometric Generalisation of Surrogate Model-Based Optimisation to Combinatorial and Program Spaces

Table 5

Fitness-distance correlation for SMs based on hamming and swap distance, after training them with 100 randomly sampled data-points.

Instance	Hamming SM	Swap SM
kra30a	0.85	0.57
kra32	0.88	0.80
lipa30a	0.93	0.39
nug30	0.88	0.71
ste36a	0.81	0.59
tho30	0.82	0.83
unih30	0.86	0.93
unis30	0.87	0.79