Daniel Falbel

Ie. use pre-trained weights.

Feedback on the API

1

I wrote a draft vignette on what I think would be a reasonable API for tft and would like to get some feedback on it. Ideally I'd love to have...

A recent blog post in google AI has the following: https://ai.googleblog.com/2021/12/interpretable-deep-learning-for-time.html

Metrics wishlist

5

- [x] AUC - [ ] Precision - [ ] Recall - [x] MSE - [x] RMSE - [x] MAE - [ ] Cohen's Kappa

Weights

1

We should provide a way for users to provide case weights as arguments for the loss function.

THis callback will trace the model for training and validation. Hopefully the optimizations that happen on tracing can generate some speedups for training more complicated models. TODO: - [ ]...

Callbacks wishlist

7

- [x] Model checkpoint callback - [x] Early stopping - [x] Learning rate schedulers - [x] ndjson/csv logger - [ ] Gradient clipping