Algorithm 1: GNBR Training Algorithm

Step 1:  Initialize the target parameters $\alpha$$, \beta$$, \mathrm{and} W_j$. Step 2:  Use the one-step LM algorithm to minimize the target network performance scale function $\mathrm{E}$. Step 3:  Approximate $\mathbf{G}$ by means of the Gaussian–Newton method and calculate the number of effective weights $\gamma$. Step 4:  Update target parameters $\alpha$ and $\beta$ with (8). Step 5:  Repeat Steps 2–4 until the values of $\alpha$ and $\beta$ converge.