Table 6 Hyper-parameters settings of SMFF-DTA

\(D_{d1}, \ D_{d2}\)

64, 128

\(D_{t1}, \ D_{t2}, \ D_{t3}\)

64, 6, 15

Learning rate

[1e-6, 5e-6, 8e-6, 1e-5](Davis), 2e-5(KIBA)

Batch size

[32, 64, 128, 256](Davis), 128(KIBA)

Interaction Block Head Number

[2, 4, 8, 16]

Cross Attention Head Number

[2, 4, 8, 16]