Skip to main content

Table 1 Effects of weights on the performance of the DCNN

From: A denoising-classification neural network for power transformer protection

Weight

Training

Validation

Test

ACC (%)

RecLoss

ClaLoss

ACC (%)

RecLoss

ClaLoss

ACC (%)

0.2/1.0

99.84

0.4279

0.7798

99.52

0.2655

9.7790

93.27

0.4/1.0

99.88

0.3967

1.7331

99.84

0.2299

2.8638

96.41

0.6/1.0

99.61

9.7129

2.5993

99.20

4.4010

6.6541

93.27

0.8/1.0

96.45

9.7129

11.0904

97.60

4.4010

5.4481

92.83

1.0/1.0

99.45

9.7129

3.6392

99.36

4.4010

7.4553

95.96

1.1/1.0

99.84

0.3074

1.7329

99.68

0.1914

1.7233

95.07

1.2/1.0

99.84

9.7129

1.4730

99.36

4.4010

3.8111

93.27

1.3/1.0

100.00

0.3236

0.6932

99.68

0.2155

13.9073

92.83

1.4/1.0

99.80

9.7129

1.9064

99.52

4.4010

0.6932

95.07

1.5/1.0

99.96

0.2226

0.2600

99.68

0.1702

5.0990

96.86

1.6/1.0

99.41

0.2357

2.7451

99.36

0.1744

23.6253

93.27

1.7/1.0

99.92

0.9992

0.9991

99.84

0.1803

0.6478

96.41

1.8/1.0

99.80

9.7129

2.8593

99.20

4.4010

12.0977

93.72

1.9/1.0

99.92

9.7129

1.0397

99.52

4.4010

6.9627

95.07

2.0/1.0

99.61

9.7129

1.2131

99.84

4.4010

4.5306

93.27