search

Hanjun-Dai / graph_adversarial_attack

Adversarial Attack on Graph Structured Data (https://arxiv.org/abs/1806.02371)
MIT License
126 stars 32 forks source link

Question about hierarchical Q-Learning #6

Open matouk98 opened 3 years ago

matouk98 commented 3 years ago

Hi, hanjun. Thanks a lot for your great work! I have a question about the hierarchical Q-Learning mentioned in the paper. In equation 11, there are 2M Q functions and the paper claims only two distinct parametrizations are enough. However, when I read the code /node_attack/q_net_node.py, in line 163, there initialize num_steps Q networks which are different from the paper. Does my understanding have any mistakes? Looking forward to your reply!!

Hanjun-Dai commented 3 years ago

Hi there,

Thanks for your interest in our work. As in our paper we only add/delete one edge, so it is equivalent.

cxw-droid commented 1 year ago

Hi,

I can see adding one edge in the code https://github.com/Hanjun-Dai/graph_adversarial_attack/blob/f2aaad73efd142bcc20d5e8c43117e5359f9aa8e/code/graph_attack/rl_common.py#L77 But I have not found where an edge is removed. Or the code did not implement edge removing? Thanks!

Hanjun-Dai commented 1 year ago

Hi, you can check the folder for node attack case.

cxw-droid commented 1 year ago

Thank you for your reply. For node attack, I only find adding edges: https://github.com/Hanjun-Dai/graph_adversarial_attack/blob/f2aaad73efd142bcc20d5e8c43117e5359f9aa8e/code/node_attack/node_attack_common.py#L33