search

zalandoresearch / pytorch-ts

PyTorch based Probabilistic Time Series forecasting framework based on GluonTS backend
MIT License
1.21k stars 191 forks source link

How can TimeGrad be used for univariate dataset? #124

Closed BEbillionaireUSD closed 1 year ago

BEbillionaireUSD commented 1 year ago

Currently, it seems can only used for multi-variate time series as the Conv1d is conducted on the "metric dim". How can I fix this issue? Thanks!

kashif commented 1 year ago

right, I believe for learning the conditional 1-dim distributions there are other choices. If you do not want to assume a parametric distribution class you can use e.g. conditional quantile spline... etc.

I do not think it makes sense to use diffusion for the 1d case... or am i mistaken?

BEbillionaireUSD commented 1 year ago

Yes, I understand. I just want to test different SOTA approaches. TimeGrad works well on multi-variate time series. Thanks for your reply!