Algorithm 1 NAS-HRIS Search Encoder for High-Resolution Remote Sensing Image Segmention

Require:$D_{train}$: the training set; $D_{valid}$: the validation set; n: batch size; initialized operation set P;

Ensure:

1: initialized the architecture variable $\alpha$ and the weights $\omega$ randomly, learning rate $\xi$, search epochs

2: repeat

3:    Sample batch of data $D_t$ from $D_{train}$;

4:    compute $L_{train}(\omega ,\alpha )-D_t$;

5:    Updata $\omega$ by gradient descent:

$\omega =\omega -\xi {\nabla }_{\omega }{L}_{train}(\omega ,\alpha )$;

6:    Sample batch of data $D_v$ from $D_{valid}$;

7:    compute $L_{valid}(\omega ,\alpha )-D_v$;

8:    Updata $\omega$ by gradient descent:

$\alpha =\alpha -\xi {\nabla }_{\alpha }{L}_{valid}(\omega -\xi {\nabla }_{\omega }{L}_{train}(\omega ,\alpha ),\alpha )$;

9: until converge