Skip to main content

Table 2 The MMAE of tested models on the test data. Architecture similar to the D–GEX with TAAFs [8] is the T–D–GEX with 1 tower, the overall best model is shown in bold

From: On tower and checkerboard neural network architectures for gene expression inference

# Towers

T–D–GEX

C–D–GEX

TR–D–GEX

CR–D–GEX

1

0.131301

0.130796

0.129107

0.129163

2

0.130187

0.130281

0.128623

0.128437

3

0.129839

0.129969

0.128548

0.128187

4

0.129735

0.129872

0.128568

0.128078

5

0.129729

0.129883

0.128617

0.128053

6

0.129707

0.129799

0.128677

0.128053

8

0.129760

0.129874

0.128864

0.128095

10

0.129804

0.129881

0.129078

0.128218

12

0.129889

0.129891

0.129291

0.128353

D–GEX (1×9,000, tanh) [7, 8]

0.163684

D–GEX with TAAFs (3×9,000,TAAFo sigmoid) [8]

0.134015