Skip to main content

Table 1 Different network architectures of DeepEnhancer

From: Predicting enhancers with deep convolutional neural networks

Layer ID

Layer Type

Size

Output shape

0

Input

–

4x1x300

1

Conv

128x4x1x8

128x1x293

2

Batchnorm

–

128x1x293

3

Conv

128x128x1x8

128x1x286

4

Batchnorm

–

128x1x286

5

Maxpooling

1 × 2

128x1x143

6

Conv

64x128x1x3

64x1x141

7

Batchnorm

–

64x1x141

8

Conv

64x64x1x3

64x1x139

9

Batchnorm

–

64x1x139

10

Maxpooling

1 × 2

64x1x69

11

Dense

256

256

12

Dropout

–

256

13

Dense

128

128

14

Softmax

2

2

  1. The size column records the convolutional kernel size, the max-pooling window size and the fully connected layer size. The output shape depicts the change of data’s shape in the flow