Network Dismantling on Signed Network by Evolutionary Deep Reinforcement Learning

. 2024 Dec 16;24(24):8026. doi: 10.3390/s24248026

Algorithm 4 DRL (deep reinforcement learning) Algorithm

Input:
Cache pool $B$ , batch size $n_{b}$ , population $P {(q)}_{E A}$ .
Output:
Population $P {(q)}_{D R L}$ .
1:
Randomly retrieve $C = {(S_{t}, a_{t}, r_{t}, S_{t + 1})}_{n_{b}}$ of size $n_{b}$ from the cache pool $B$ ;
2:
Calculating $L (W^{1}, W^{2})$ by Equation (17);
3:
Calculating $Δ (W^{1}, W^{2})$ of $L (W^{1}, W^{2})$ by Equation (18);
4:
Based on $Δ (W^{1}, W^{2})$ , reverse update $(W^{1}, W^{2})$ corresponding to the most favorable population $P {(q)}_{E A}$ , thus obtaining a new population $P {(q)}_{D R L}$ ;