NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:5182
Title:A Unified Framework for Data Poisoning Attack to Graph-based Semi-supervised Learning

I recommend acceptance of this article subject to some corrections. The novelty of considering data poisoning in GSSL can be interesting for the NeurIPS community. The results presented here are founded by theoretical results and experiments are conducted. It opens new lines of research and can trigger new results. Because the paper has several weak points, we insist on the importance to handle the following comments. 1/ The quality of writing is not at the expected level and some improvements have to be done before publishing. However, the overall structure of the presentation is acceptable. Correction of typos and a better explanation of the notations can be done. This issue could be solved without drastically changing the paper. 2/ The motivations behind data poisoning in GSSL was not clearly exposed. Moreover, the paper has not proposed counter-measures as it would be expected. Some elements about the former were given in the rebuttal. 3/ The related work has no deep exposition of previous results on robustness and stability of GSSL methods that could be useful to understand and compare with the work done in this paper.