. 2024 Feb 19;14:4104. doi: 10.1038/s41598-024-53964-7

Table 2.

The training algorithm flow of the SAE-HFCM model.

Algorithm: SAE-HFCM model training algorithm
Input: $ε$ : training accuracy of SAE-FCM; $\{X, Y\}$ : training data; Output: $W_{1}$ , $W_{2}$ , $W_{3}$ , $W_{x}$ , $b_{1}$ and $b_{2}$
1: Get $Z$ from input data $X$ with SAE, and record $W_{1}$ and $b_{1}$
2: Randomly initialize $W_{2}$ , $W_{x}$ and $b_{2}$ , and get $Z$ from $H$
3: Calculate $W_{3}$ by ridge regression
4: Fix $W_{1}$ , $b_{1}$ and $W_{3}$ , and update $W_{2}$ , $W_{x}$ and $b_{2}$
5: while Training accuracy < $ε$ $□$ do <
6: Update $W_{1}$ and $b_{1}$
7: Update $W_{2}$ , $W_{x}$ and $b_{2}$
8: Update $W_{3}$
9: end while
10: return $W_{1}$ , $W_{2}$ , $W_{3}$ , $W_{x}$ , $b_{1}$ and $b_{2}$