TABLE 2.

Summary of the results. The baselines of the three submitted methods are in boldface.

	Operator	Localization	Divide	Clusters	Fine-tuning	Result (km)
Greedy	—	—	—	—	—	129,775
UEF and DoLoWire:
UEF (no tuning)	Rand mix	Clusters	K-means	12,764	—	124,162
UEF	Rand mix	Clusters	K-means	12,764	Two clusters	122,226
UEF (grid clusters)	Rand mix	Clusters	Grid	1,268	Two clusters	147,817
DoLoWire (no tuning)	2- to 3-opt	Grid	Grid	1,268	—	120,751
DoLoWire	2- to 3-opt	Grid	Grid	1,268	500 points	115,620
LKH with clustering:
LKH (k-means)	2- to 5-opt	Neighbors	K-means	12,764	—	121,020
LKH (k-means)	2- to 5-opt	Neighbors	K-means	1,268	—	117,496
LKH (grid)	2- to 5-opt	Neighbors	Grid	1,268	—	113,598
LKH (grid + tuning)	2- to 5-opt	Neighbors	Grid	1,268	500 points	112,404
LKH (grid + tuning)	2- to 5-opt	Neighbors	Grid	1,268	Two clusters	111,636
LKH without clustering:
LKH	2- to 5-opt	Neighbors	—	—	—	109,284
LKH (Delaunay)	2- to 5-opt	Delaunay	—	—	—	108,996