Table: Number of function-evaluations performed on the testset of Dixon and -Szegö

programs

S5

S7

S10

H3

H6

GP

BR

C6

SHU

Bremmerman

(a)

(a)

(a)

(a)

(a)

(a)

250

 

 

Mod. Bremmerman

(a)

(a)

(a)

(a)

515

300

160

 

 

Zilinskas

(a)

(a)

(a)

8641

 

 

5129

 

 

Gomulka-Branin

5500

5020

4860

 

 

 

 

 

 

Törn

3679

3606

3874

2584

3447

2499

1558

 

 

Gomulka-Törn

6654

6084

6144

 

 

 

 

 

 

Gomulka-V.M.

7085

6684

7352

6766

11125

1495

1318

 

 

Price

3800

4900

4400

2400

7600

2500

1800

 

 

Fagiuoli

2514

2519

2518

513

2916

158

1600

 

 

De Biase-Frontini

620

788

1160

732

807

378

587

 

 

Mockus

1174

1279

1209

513

1232

362

189

 

 

Bélisle et al. (b)

 

 

 

339

302

4728

1846

 

 

Boender et al.

567

624

755

235

462

398

235

 

 

Snyman-Fatti

845

799

920

365

517

474

 

178

 

Kostrowicki-Piela

-

-

-

200

200

120

 

120

 

Yao

 

 

 

 

 

 

1132

<6000

 

Perttunen

516

371

250

264

 

82

97

54

197

Perttunen-Stuckman

109

109

109

140

175

113

109

96

(a)

Jonet et al.

155

145

145

199

571

191

195

285

2967

Storn-Price (c)

6400

6194

6251

476

7220

1018

1190

416

1371

MCS(e)

84

129

103

79

111

81

41

42

69

ASA(d)

-

-

604

74

599

197

332

269

343

ASA

-

-

604

74

599

197

332

269

343

VFSR (d)

-

-

-

133

854

406

364

264

287

VFSR

-

-

-

133

802

406

364

264

287

Toms667

-

-

1557

4461

3109

3243

1509

743

-

Csendes(d)

779

1082

2515

123

316

133

115

121

171

Csendes

-

-

-

123

316

133

115

121

171

Csendes (e)

-

-

-

-

-

-

264

129

-

Genocop 3.0

-

-

-

2621

-

1327

1952

1535

2792

Genoocpp III

-

-

-

4229

-

5130

9659

5190

-

DE

8378

7661

8374

1926

-

1819

4534

2598

11589

PGA

-

-

-

1898

-

10836

-

1313

2426

random

-

-

-

-

-

-

-

-

-

(a) No convergence to a global Minimum.

(b) Average of the several runs. In case of the function H6 only 70% of the runs converge.

(c) Average of 25 runs. In case of H6 only 24 runs
(once there was no convergence within 12000 evaluations).

(d) Implemented stoping-rules were evaded.

(e) Local search was performed using random numbers.

- The required accuracy wasn't reached after 12000 evaluations.