Algorithm 1 The algorithm of FFCNN back-propagation.

Input: Labeled source domain samples ${\left\{({\mathit{x}}_i^S,{\mathit{y}}_i)\right\}}_{i=1}^m$, unlabeled target domain samples

${\left\{\left({\mathit{x}}_i^t\right)\right\}}_{i=1}^m$, regularization parameter $\lambda$, learning rate $\eta$, dilate rate $\{r_1,r_2,r_3\}$.

Output Network parameters $\left\{{\theta }_{r_j}^{conv1},{\theta }^{conv2},{\theta }^{fc},{\theta }^{clf}\right\}$ and predicted labels for target

domain samples.

Begin:

Initialization for $\left\{{\theta }_{r_j}^{conv1},{\theta }^{conv2},{\theta }^{fc},{\theta }^{clf}\right\}$.

while stopping criteria is not met do

for each source and target domain samples of mini-batch size $m^{\prime }$ do

Calculate output ${\mathit{x}}_{r_j}^{conv1}$ of each branch in dilate convolution layer according to

Equation (9).

Connect ${\left\{{\mathit{x}}_{r_j}^{conv1}\right\}}_{j=1}^3$, and calculate output of the second convolution layer according

to Equation (10).

Calculate features representations ${\mathit{z}}_i$ and output of softmax layer according to

Equation (11).

Calculate loss $\ell ({\mathit{y}}^S,{\tilde{\mathit{y}}}^S,{\mathit{z}}^S,{\mathit{z}}^T)$ according to Equation (12)

Upgrade $\left\{{\theta }_{r_j}^{conv1},{\theta }^{conv2},{\theta }^{fc},{\theta }^{clf}\right\}$ according to Equation (13).

end for

end while