Skip to main content

Table 2 Model selection results. Six different hyperparameter sets for final training of DBN

From: Unsupervised deep learning reveals prognostically relevant subtypes of glioblastoma

Set ID

Hidden Layer Sizes

Learning Rates

Epochs

Input Size

1st

2nd

3rd

pretrain

train

pretrain

train

1

1100

500

100

7.75E-05

2.41E-03

14

101

9476

2

1300

700

100

7.75E-05

2.41E-03

14

91

9476

3

1300

700

150

7.75E-05

2.41E-03

14

88

9476

4

1300

700

150

7.75E-05

2.41E-03

14

62

9476

5

1400

1000

200

3.03E-03

3.26E-03

14

40

7160

6

1300

700

150

7.75E-05

2.41E-03

14

97

7160