AUGUST 2022 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE 41 TABLE II Detailed settings of the large-scale sparse MOPs used in experiments. BENCHMARK PROBLEM TYPE OF VARIABLES SMOP1-SMOP8 Real NO. OF VARIABLES 100 500 1000 5000 Neural network training NN1 NN2 NN3 NN4 Feature selection FS1 FS2 FS3 FS4 Pattern mining PM1 PM2 PM3 PM4 Community detection CD1 CD2 CD3 CD4 Critical node detection CN1 CN2 Type of variables Binary Type of variables Binary Type of variables Real No. of variables 301 521 1241 6241 No. of variables 100 500 800 4434 Type of variables Binary No. of variables 100 500 1000 5000 Type of variables Binary No. of variables 105 453 1133 4039 No. of variables 102 500 Dataset Wine1 Statlog(German)1 Connectionist Bench Sonar1 LSVT Voice Rehabilitation1 Dataset Hill_Valley1 Madelon1 Gse725262 GLIOMA3 Dataset Synthetic [50] Synthetic [50] Synthetic [50] Synthetic [50] Dataset polbooks4 celegans_metabolic5 Email5 Facebook [51] Dataset Movies6 FF500 No. of samples 178 1000 208 126 No. of samples 606 2600 61 50 No. of transactions 1000 5000 10000 50000 No. of nodes 105 453 1133 4039 No. of nodes 102 500 No. of features 13 24 60 310 No. of features 100 500 800 4434 No. of items 100 500 1000 5000 No. of edges 441 4596 5451 88234 No. of edges 243 1078 No. of classes 3 2 2 2 No. of classes 2 2 4 4 Avg. length of transactions 50 250 500 2500 SPARSITY OF PARETO OPTIMAL SOLUTIONS 0.1 NO. OF OBJECTIVES 2