Benchmark for the k-cut problem
This page compares the performance of several heuristics, which can be downloaded from this page.
The evaluated instances are hypergraphs with 100 to 400 vertices and 500 to 1000 hyperedges. These instances are available here.
The heuristics were tested for two values of k: k = 10 and k = 50. The tables below present a side-by-side comparison of their performance, including solution cost and computation time.
Detailed results for k = 10 are available here, and for k = 50, they are available here.
Solution Cost | CPU (in seconds) | |||||||
---|---|---|---|---|---|---|---|---|
Heuristic | Fast | GHT | Hyb | Pack | Fast | GHT | Hyb | Pack |
Instance | ||||||||
h_100_1000 | 965 | 961 | 961 | 961 | 0.001 | 0.098 | 0.100 | 0.053 |
h_100_500 | 438 | 436 | 436 | 438 | 0.001 | 0.050 | 0.075 | 0.025 |
h_200_1000 | 439 | 439 | 439 | 439 | 0.002 | 0.200 | 0.300 | 0.100 |
h_200_500 | 129 | 129 | 129 | 129 | 0.001 | 0.100 | 0.500 | 0.300 |
h_400_1000 | 124 | 124 | 124 | 124 | 0.003 | 0.400 | 1.900 | 1.500 |
h_400_500 | 41 | 40 | 40 | 40 | 0.002 | 0.200 | 5.700 | 5.400 |
Solution Cost | CPU (in seconds) | |||||||
---|---|---|---|---|---|---|---|---|
Heuristic | Fast | GHT | Hyb | Pack | Fast | GHT | Hyb | Pack |
Instance | ||||||||
h_100_1000 | 4020 | 4066 | 4066 | 4066 | 0.011 | 0.100 | 0.200 | 0.060 |
h_100_500 | 1897 | 1878 | 1878 | 1878 | 0.003 | 0.054 | 0.079 | 0.028 |
h_200_1000 | 2236 | 2236 | 2236 | 2223 | 0.010 | 0.200 | 0.300 | 0.100 |
h_200_500 | 951 | 947 | 941 | 947 | 0.004 | 0.100 | 0.500 | 0.400 |
h_400_1000 | 921 | 909 | 900 | 909 | 0.008 | 0.400 | 1.900 | 1.500 |
h_400_500 | 344 | 341 | 347 | 342 | 0.004 | 0.300 | 6.600 | 6.400 |