Great presentation! Am I correct in thinking that the generator also acts as a decoder? Also, what is the purpose of having Cz? I don’t understand what the benefit is of distinguishing noise from embeddings.
i think the Cz distinguishes the random noise and encoded time series so that the function of encoder can be served as a feature extractor. And the purpose is to let the Z distripution be as close as possible to X(original time series) distripution. if i'm wrong, please rectify me.
Whether we can do it for Univariate Time series Model?
How many epochs was the model trained for for each dataset out of curiosity? Thank you ! Love the paper!
I plan to start in this direction about anomaly detection. How about the results in high dimensional time series。
哇，讲的太棒了。你的论文我去年就看了。已关注你了。
Very nice presentation, thanks a lot
how about no stationary time series?
Great presentation, really well put together, thanks!
Thank you!
Great presentation! Am I correct in thinking that the generator also acts as a decoder?
Also, what is the purpose of having Cz? I don’t understand what the benefit is of distinguishing noise from embeddings.
i think the Cz distinguishes the random noise and encoded time series so that the function of encoder can be served as a feature extractor. And the purpose is to let the Z distripution be as close as possible to X(original time series) distripution. if i'm wrong, please rectify me.
Exactly I believe Lieon offered an excellent explanation here.
wondefull presentation. How to code this??
Hi here is the details about how we use tensorflow+keras to implement it github.com/signals-dev/Orion/blob/master/orion/primitives/tadgan.py