search

ML4ITS / mtad-gat-pytorch

PyTorch implementation of MTAD-GAT (Multivariate Time-Series Anomaly Detection via Graph Attention Networks) by Zhao et. al (2020, https://arxiv.org/abs/2009.02040).
MIT License
328 stars 76 forks source link

Some question about the param target_dims #11

Closed ShaoSK closed 2 years ago

ShaoSK commented 3 years ago

I have the question about this param, as the mtad-gat is a multivariable time series model which uses modules to catch the time dependencyand the feature dependency, if use only one dim to training and testing, it just degenerates into univariate time series model. What's the use of the corresponding module of feature dependency in this situation?

axeloh commented 2 years ago

Hello,

MTAD-GAT is indeed a model made for multivariate time-series. If the input is univariate, then the feature-oriented GAT module will have no effect. However, the time-oriented GAT module will still serve the purpose of attending to different time-steps.

Recall that even though datasets SMAP and MSL have univariate output, the input is still multivariate, and so both GAT modules will be utilized.