mtad_gat

MTAD-GAT

Citekey	ZhaoEtAl2020Multivariate
Source Code	own
Learning type	semi-supervised
Input dimensionality	multivariate

Dependencies

• python >= 3.7
• torch

Notes

MTAD-GAT outputs anomaly scores for windows. The results require post-processing. The scores for each point can be assigned by aggregating the anomaly scores for each window the point is included in.

You can use the following code snippet for the post-processing step in TimeEval (default parameters directly filled in from the source code):

from timeeval.utils.window import ReverseWindowing
# post-processing for MTAD-GAT
def post_mtad_gat(scores: np.ndarray, args: dict) -> np.ndarray:
    window_size = args.get("hyper_params", {}).get("window_size", 20)
    return ReverseWindowing(window_size=window_size + 1).fit_transform(scores)

SR

This algorithm uses Spectral Residuals to clean training data. We used the code from this algorithm.