sktime / sktime/pytorch-forecasting
Hybrid Forecast Model with TFT
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 5k
- Forks
- 912
- Avg merge
- 3d 13h
- Merged PRs (30d)
- 12
Description
Hi,
First of all thank you very much for this remarkable piece of work!
I don't know whether it is supported but there are use cases where forecasting time varying reals and time varying categories at the same times turns useful. For example predicting next events in a sequence and their durations.
How would such hybrid targets be handled in terms of loss and say target_normalizer? Shall we pass lists such as for example:
loss= [CrossEntropy(), QuantileLoss()]? Didn't seem to work on first attempt.
Many thanks in advance!
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reviewing the TFT target handling around loss and target_normalizer; the issue names no files or tests. Determine whether simultaneous time-varying real and categorical targets are supported and define the expected loss and normalizer interface before adding coverage.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100