Skip to main content

Table 2 Different hyperparameter settings for 2D-Vanilla-CNN model based on the trained and tested statistical measures. The final selected parameters are highlighted

From: Convolutional neural network models for cancer type prediction based on gene expression

Hyperparameters

Loss

dense layer size

filter

kernel

stride

mean train_score

stdev train_score

mean test_score

stdev test_score

128

32

(7, 7)

(1, 1)

20.999

18.228

21.281

14.904

128

32

(7, 7)

(2, 2)

0.005

0.002

0.192

0.022

128

32

(10, 10)

(1, 1)

21.398

18.582

21.771

15.298

128

32

(10, 10)

(2, 2)

0.009

0.003

0.187

0.008

128

32

(20, 20)

(1, 1)

0.027

0.004

0.202

0.029

128

32

(20, 20)

(2, 2)

0.043

0.011

0.206

0.009

128

64

(7, 7)

(1, 1)

10.213

17.688

10.566

14.618

128

64

(7, 7)

(2, 2)

0.004

0.001

0.187

0.018

128

64

(10, 10)

(1, 1)

31.430

1.149

31.675

1.019

128

64

(10, 10)

(2, 2)

0.012

0.006

0.177

0.014

128

64

(20, 20)

(1, 1)

12.020

18.052

12.149

14.818

128

64

(20, 20)

(2, 2)

0.055

0.016

0.204

0.020

512

32

(7, 7)

(1, 1)

21.245

18.419

21.175

14.815

512

32

(7, 7)

(2, 2)

10.944

18.953

11.022

15.306

512

32

(10, 10)

(1, 1)

10.964

18.987

11.148

15.482

512

32

(10, 10)

(2, 2)

0.003

0.001

0.213

0.025

512

32

(20, 20)

(1, 1)

10.988

19.002

11.132

15.436

512

32

(20, 20)

(2, 2)

1.110

1.849

1.271

1.397

512

64

(7, 7)

(1, 1)

31.430

1.149

31.675

1.019

512

64

(7, 7)

(2, 2)

10.213

17.688

10.560

14.622

512

64

(10, 10)

(1, 1)

31.497

1.211

31.648

1.087

512

64

(10, 10)

(2, 2)

20.628

17.858

20.481

14.363

512

64

(20, 20)

(1, 1)

11.299

16.825

11.562

13.969

512

64

(20, 20)

(2, 2)

12.020

18.046

12.152

14.776