MSCRED performance

[DaveFantini]

Hi,
I was reading about multivariate time series anomaly detection algorithm and the paper seemed interesting to me, can you provide some tests results which shows the actual "bullshit" performance you mentioned in the README ?
The the artificial data generated with the simulate_data.py was your only experiment ?

[sangramkapre]

Hi, were you able to successfully run the model on any data? I was curious to know if this model really works. The author of this repo has mentioned that he couldn't get the results claimed in the paper though he had implemented everything to the last detail. What was your experience?

[DaveFantini]

Hi, I have implemented the algorithm without the attention part. I need to test on real dataset yet... It is difficult to find multivariate time series with labels.. I'll let you know if I find something interesting :) Il 16 giu 2020 01:10, Sangram Kapre <[email protected]> ha scritto: Hi, were you able to successfully run the model on any data? I was curious to know if this model really works. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#1 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AGMBFDKNNFFHMVTJVPXASLTRW2S7BANCNFSM4KH2ORGQ>.

[sangramkapre]

Thanks! Also, did you get reasonable results on the synthetic dataset provided in the repo? I am going through that right now. Looking forward to knowing more about your efforts soon!

[PeterA182]

I'm curious on this as well. Please do tell if anyone finishes up an application of it. Trying to spend time on it myself.

[neozhao98]

Now, I can't agree with you @albertwujj more

[yasirroni]

Now, I can't agree with you @albertwujj more

Hi @neozhao98, you means to disagree that the paper is bullshit right? Can you explain and provide code implementation that really work and tested, both using the author data and outside data?

If we look to the author repo, no confirmation from user that the code, implementation, or model that works.

Edit:

My bad, "I can't agree with you more" have same meaning to "I can't more agree with you" (that means you are agree with this author repo that the paper is bullshit).

So, it's more and more confirmed that the paper is a sscam.