search

squareRoot3 / Rethinking-Anomaly-Detection

"Rethinking Graph Neural Networks for Anomaly Detection" in ICML 2022
https://proceedings.mlr.press/v162/tang22b.html
176 stars 29 forks source link

Problem regarding your research paper #6

Closed cankaili121 closed 1 year ago

cankaili121 commented 1 year ago

Dear Tang,

I am writing to express my appreciation for the work that you and your team have done on your recent research paper. However, after reading the paper, I was unable to find a clear research "problem". While I found your “observations” regarding the phenomenon of right-shifted spectral energy to be thorough and well-explained, I am confused as to what ”problems“ this phenomenon can cause. Does the right-shifted spectral energy lead to a decrease in anomaly detection accuracy? The paper does not seem to provide a clear illustration or explanation.

Additionally, I am curious about the necessity of the spectral graph convolutional network that you designed. Is it necessary for solving the research "problem"? I would appreciate any insights or explanations that you can provide on this matter.

Thank you for your time and effort in conducting this research.

sqc-cser commented 1 year ago

if you are a Chinese, I will recommend you to watch his video about this. https://www.bilibili.com/video/BV1BG411p7og/

squareRoot3 commented 1 year ago

Thanks for the response of sqc-cser. We also have a short English talk in https://slideslive.com/38984068/rethinking-graph-neural-networks-for-anomaly-detection?ref=speaker-24150 I think our paper and talk have clearly addressed your question.