Algorithm 2. An overview of P2P-KD for Domain-IL

1. $t=2$

2. while true do

3. Input: previous model ${\mathcal{M}}_{t-1}$ , training dataset $D_t$ , other hyper-parameters for

4. Algorithm 1: $p,k,\alpha$

5. Output: model ${\mathcal{M}}_t$

6. Initialize ${\mathcal{M}}_t={\mathcal{M}}_{t-1}$

7. Freeze ${\mathcal{M}}_{t-1}$

8. while true do

9. Sample a batch dataset D from $D_t$

10. Get $D_{trans}$ , $D_{pri}$ , $D_{pub}$ using Algorithm 1

11. $D_{pub}=D_{pub}\cup D_{trans}$

12. Train ${\mathcal{M}}_t$ by minimizing Equation (3)

13. end while

14. $t=t+1$

15. end while