Skip to main content

Table 8 Results of the Pdrop sweep using the best setting of the double-task network (C2-LSTM, L=6,N=256)

From: Exploring convolutional, recurrent, and hybrid deep neural networks for speech and music detection in a large audio dataset

P drop

Train

Validation

Test

 

Cost

Acc.%

Cost

Acc.%

Cost

Acc.%

0

0.651

74.43

0.726

71.48

0.733

70.98

0.1

0.634

75.02

0.736

71.25

0.745

70.25

0.2

0.669

73.69

0.708

72.19

0.721

71.25

0.3

0.601

76.10

0.704

73.01

0.721

72.10

0.4

0.668

73.37

0.692

73.43

0.701

72.44

0.5

0.702

72.12

0.726

71.51

0.741

70.91

0.6

0.734

71.06

0.723

71.78

0.735

70.91

  1. The best setting in terms of validation cost is highlighted in italics