Algorithm 1 Auto Graph Self-supervised strategy optimization

Input: $G=\{V,E\}$

Parameters: hyperparameter optimization iterations $N_{select}$, $\beta$ fine-tuning iterations $N_{\beta }$,

labeled set $\mathcal{L}$

Initialization: ${\theta }^0$, $w^0$, $W_s^0$, $B^0$, ${\mathrm{\Lambda }}^0$

divide $\mathcal{L}$ into 3 similar parts, ${\mathcal{L}}_0\cup {\mathcal{L}}_1\cup {\mathcal{L}}_2=\mathcal{L}$

while $l\in \{{\mathcal{L}}_0,{\mathcal{L}}_1,{\mathcal{L}}_2\}$ do

Validation set: l, training set: $\mathcal{L}\setminus l$

repeat

$n\leftarrow n-1$

Update $({\theta }^n,w^n,w_s^n)$ based on Equation (8)

Update $(B^n,{\mathrm{\Lambda }}^n)$ based on Equation (15)

until $n\ge N_{select}$

end while

$i\leftarrow argmax\mathrm{\Lambda }$, set ${\lambda }_i\leftarrow 1$, ${\lambda }_n\leftarrow 0(n\ne i)$

while $l\in \{{\mathcal{L}}_0,{\mathcal{L}}_1,{\mathcal{L}}_2\}$ do

Validation set: l, training set: $\mathcal{L}\setminus l$

repeat

$n\leftarrow n-1$

Update $({\theta }^n,w^n,w_s^n)$ based on Equation (8)

Update $B^n$ based on Equation (15)

until $n\ge N_{\beta }$

end while

Output: $\theta ,w,W_s,B,\mathrm{\Lambda }$