Table 2 The performance comparison of different encoding methods and different hyperparameter configurations.

From: Quantum integration in swin transformer mitigates overfitting in breast cancer screening

 

ACC

BACC

AUC

SEN

SPE

Encoding

Angle embedding

0.902

0.904

0.960

0.933

0.865

Amplitude embedding

0.885

0.888

0.956

0.927

0.837

Number of variational layer

1

0.898

0.901

0.961

0.935

0.855

2

0.902

0.904

0.960

0.933

0.865

3

0.900

0.902

0.958

0.932

0.862

4

0.900

0.902

0.959

0.932

0.862

5

0.896

0.898

0.955

0.932

0.855

Variational gates

U3+CZ

0.902

0.904

0.960

0.933

0.865

U3+CNOT

0.887

0.892

0.953

0.940

0.829

RY+CZ

0.894

0.896

0.957

0.926

0.857

Optimizer

SGD

0.902

0.904

0.960

0.933

0.865

Adam

0.865

0.872

0.941

0.931

0.799

AdamW

0.875

0.881

0.944

0.936

0.813

Back to article page