Giters

KDD-OpenSource / DeepADoTS

Repository of the paper "A Systematic Evaluation of Deep Anomaly Detection Methods for Time Series".

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Autoencoder Algorithm

trungkien210493 opened this issue · comments

commented

Dear @WGierke and team,
According to your reference "Outlier Detection Using Replicator Neural Networks" , DaWaK 2002, I see that in the paper, they implemented the RNN with 5 layers. But in your code "src/algorithms/autoencoder.py", you implement a simple auto-encoder, that mean it contains only 4 layers. Then, I think you should to correct the reference in "readme" file. On other hand, to detect the anomalies, you use the multivariate normal random variable, it is nearly the same ideal of "LSTM-based Encoder-Decoder for Multi-sensor Anomaly Detection", but I'm not clear why you rolling the data by using the "sequence_length" variable and calculate the "mean" of the score according to the rolling data. Anyway, this is a good repos to anyone to learning and testing deep learning.