Ultrasound Evaluation of the Primary α Phase Grain Size Based on Generative Adversarial Network

. 2022 Apr 24;22(9):3274. doi: 10.3390/s22093274

Algorithm 1: Effective virtual sample generation algorithm ( $EVSG$ algorithm)
Input: real training sample set $D_{s}$
Output: optimal virtual sample set $D_{v i r}^{″}$
1.	begin
2.	/* initialize the number of iterations $i$ , initialize reconstructed sample set $D_{f} (0)$ , initialize $N_{v i r}$ , initialize the virtual sample set after initial screening $D_{v i r}^{'}$ */
3.	$i \leftarrow 0$
4.	$D_{f} (0) \leftarrow D_{s}$
5.	$N_{v i r} \leftarrow 1$
6.	$D_{v i r}^{'} \leftarrow \emptyset$
7.	repeat
8.	$i \leftarrow i + 1$
9.	repeat
10.	$D_{v i r} = (X_{v i r}, Y_{v i r}) = G (D_{s}, N_{v i r})$
11.	if $X_{v i r} \in [X_{\min}, X_{\max}]$ then
12.	$D_{v i r}^{'} = D_{v i r}^{'} \cup D_{v i r}$
13.	end if
14.	until $\| D_{v i r}^{'} \| = N$
15.	$D_{f} (i) = D_{f} (i - 1) \cup D_{v i r}^{'} (i)$
16.	until $OTR (D_{f} (i)) \leq OTR (D_{f} (i - 1))$
17.	$D_{v i r}^{″} = {\begin{cases} D_{v i r}^{'} (i - 1) \cup D_{v i r}^{'} (i - 2) \cup \dots \cup D_{v i r}^{'} (1) & i \geq 1 \\ \emptyset & i = 1 \end{cases}$
18.	end