setting	method	energy distance to the optimized measure	energy distance to the data	seconds
plain	random	0.0024	0.0024	–
plain	k-center greedy	0.0228	0.0228	1.9
plain	herding · energy	0.00107	0.00107	1.5
plain	twinning	0.00166	0.00166	0.43
plain	kernel thinning · energy	0.00122	0.00122	3.1
plain	support points · energy	0.00239	0.00239	8.6
weights = length	random	0.00504	0.0024	–
weights = length	k-center greedy	0.0336	0.0228	2
weights = length	herding · energy	0.00103	0.00439	1.5
weights = length	kernel thinning · energy	0.00107	0.00448	4.9
weights = length	support points · energy	0.00594	0.00239	8.5
reference = cs	random	0.147	0.0024	–
reference = cs	k-center greedy	0.0909	0.0228	2.1
reference = cs	herding · energy	0.00818	0.0994	0.94
reference = cs	kernel thinning · energy	0.00812	0.101	4.7
reference = cs	support points · energy	0.14	0.00242	4.6
plain, n = 250	kernel thinning · compress = :never	0.00244	0.00244	4.5
plain, n = 250	kernel thinning · compress = :always	0.00258	0.00258	3.1
plain, n = 250	random	0.00485	0.00485	–