Table 2 Average quality scores for large real networks

From: CLOVE, a Travelling Salesman’s approach to hyperbolic embeddings of complex networks with communities

 

MA

EPP

AUC

GR

GS

GE

Running Time (min.)

Peak Mem. (GB)

CLOVE(default)

0.278

0.405

0.964

0.304

0.223

0.079

663.259

2.708

CLOVE (with SA)

0.276

0.409

0.964

0.306

0.225

0.079

693.016

3.163

CLOVE (Louvain)

0.277

0.406

0.962

0.291

0.214

0.076

649.248

3.181

HMCS

0.271

0.108

0.955

0.073

0.059

0.021

89.061

3.113

  1. We display the measured average scores for the mapping accuracy, MA, the edge prediction precision, EPP, the area under the receiver operating characteristic curve, AUC, the greedy routing score, GR, the greedy success rate, GS and the greedy routing efficiency, GE, averaged over 17 real networks with size ranging between N = 20000 and N = 2.7 ⋅ 106 nodes. For each indicator, the best performing value is shown in bold. Besides the quality scores, we also display the running time and the peak memory usage in GB. In the top part of the table, we list the scores obtained for CLOVE with default settings, for CLOVE with simulated annealing optimisation during the solution of the TSP problem and for CLOVE with Louvain communities. For comparison, in the bottom row we give the results for HMCS.
Back to article page