phile
phile
Sorry I do not get your point. Our work is aim to generate synthetic data but no recover certain data.
We did not evaluate privacy-preserving metrics.
I use `--hidden-dim 112 --embed-dim 448 --noise-dim 448 --layers 3` in practice for Energy dataset. `--hidden-dim` should be 4 times the size of input features as the paper says and...
Sorry, I did not save the code for the baselines. You can adapt them in their corresponding repos. The idea is just following how we use RTSGAN to generate incomplete...
You can check the parameter setting in the paper: For stock, it is default in the code and for Energy, it should be `--hidden-dim 112 --embed-dim 448 --noise-dim 448 --layers...
It should work. You can just remove the missing data part. I use `seq_len` to control the length of time series. During the decoding we first get the generated synthetic...
Yes. Note that for `stock_energy` part, it does not include static features which would contain the intended `seq_len`. By looking to `general` part you can find that we first generate...
Does your training data contains different `seq_len`?
I think the TimeGAN repository only allows evaluations on fixed-length time series. You may need to modify their codes.
Hi, I do not save the seed and set the seed of Pytorch. The command you use is not for stock and energy datasets. You should use the following command...