Reliable Breast Cancer Diagnosis with Deep Learning: DCGAN-Driven Mammogram Synthesis and Validity Assessment

Table 3

The discriminator architecture.

Layer no	Layer type	Output shape	Activation
1	Conv2D (input)	64 × 32 × 32	LeakyReLU
2	Conv2D	64 × 16 × 16	LeakyReLU
3	Batch normalization	64 × 16 × 16	—
4	Conv2D	128 × 8 × 8	LeakyReLU
5	Batch normalization	128 × 8 × 8	—
6	Conv2D	256 × 4 × 4	LeakyReLU
7	Batch normalization	256 × 4 × 4	—
8	Conv2D	512 × 4 × 4	—
9	Batch normalization	512 × 4 × 4	LeakyReLU
10	Flatten	1 × 1 × 1	Sigmoid