Skip to main content

Table 1 The first adopted CNN network layout

From: Conditionally optimal classification based on CFAR and invariance property for blind receivers

Layer Output dimensions
Input layer 2 ×1024 ×1
Conv. 2 ×1024 ×16
Batch normalization 2 ×1024 ×16
ReLU 2 ×1024 ×16
Max pool 2 ×512 ×16
Conv. 2 ×512 ×24
Batch normalization 2 ×512 ×24
ReLU 2 ×512 ×24
Max pool 2 ×256 ×24
Conv. 2 ×256 ×32
Batch normalization 2 ×256 ×32
ReLU 2 ×256 ×32
Max pool 2 ×128 ×32
Conv. 2 ×128 ×48
Batch normalization 2 ×128 ×48
ReLU 2 ×128 ×48
Max pool 2 ×64 ×48
Conv. 2 ×64 ×64
Batch normalization 2 ×64 ×64
ReLU 2 ×64 ×64
Max pool 2 ×32 ×64
Conv. 2 ×32 ×96
Batch normalization 2 ×32 ×96
ReLU 2 ×32 ×96
Average pooling 2 ×1 ×96
Fully connected 1 ×1 ×2
Soft Max 1 ×1 ×2