Table 3 Simulation results of CNN and CWNN on MNIST.

From: Research on improved convolutional wavelet neural network

Numbers of SPs and statistical item

Total numbers of ACs

Error rate of CNN (accuracy)

MSE of CNN

Running times of CNN

Error rate of CWNN (accuracy)

MSE of CWNN

Running times of CWNN

1

6000

0.1584

0.1604

163.2180

0.1114

0.1368

217.7990

2

6000

0.1592

0.1593

164.2330

0.2453

0.2010

228.4130

3

6000

0.1578

0.1591

173.3970

0.1221

0.1250

218.7320

4

6000

0.1499

0.1532

173.3970

0.1088

0.1287

253.8580

5

6000

0.1701

0.1691

190.1890

0.1068

0.1330

244.1440

6

6000

0.1771

0.1759

185.5980

0.1425

0.1586

238.8930

7

6000

0.1599

0.1631

194.7600

0.1184

0.1386

534.7950

8

6000

0.1762

0.1747

191.8820

0.1173

0.1388

119.7400

9

6000

0.1541

0.1581

196.8820

0.1088

0.1267

282.4970

10

6000

0.1807

0.1702

237.1220

0.1101

0.1274

250.3990

Average

6000

0.16434 (0.83566)

0.16431

187.0700

0.12915(0.87085)

Best

0.14146 Smaller

258.9270

  1. Underlined values represent the results with the highest accuracy or the lowest mean square error.