Algorithm 1: Student network training

${\mathrm{B}}^{\ast }$: Soft target dataset

${\mathrm{W}}_S$: Parameters of a student network

1. Initialise ${\mathrm{W}}_S$

2. Input ${\mathrm{B}}^{\ast }$ to pretrained network

3. Compute loss function $ℒ\left({\mathrm{W}}_S\right)$ using Equation (4)

4. ${\mathrm{W}}_S^{\ast }\leftarrow \arg \underset{W_S}{\min }ℒ\left({\mathrm{W}}_S\right)$

5. Select the best parameter ${\mathrm{W}}_S^{\ast }$ for student network