ESRNN-GPU
ESRNN-GPU copied to clipboard
damitkwr
end-to-end example as a notebook?
Any chance you could push an end-to-end example as a jupyter notebook? It's really hard to follow the codebase... thanks!
I'm excited to see the end-to-end example. I'd like to generalize this to non-M4 time series, and that'll be super helpful for it. Thanks!
I'm excited to see the end-to-end example. I'd like to generalize this to non-M4 time series, and that'll be super helpful for it. Thanks!
Very interested in this
Would be good a generalization for different data and a way to get the forecasted values.
Would also be very grateful for any further documentation for the code beyond Smyl's Uber page and the paper. I've come from using Keras and finding it difficult to follow. It's training fine but would like to generalise on non-M4 data (though also not sure how to obtain forecasts). Even restricting the M4 data to a fewer number of time series breaks the code due to a batch size issue in Pytorch. Will keep trying to get my head around it. Many thanks.
Why has nothing been done to this end? Some smart person somewhere should have a colab they can share.
Sure, we will start on getting a notebook example out
Hi, Sir, do you have update on this one. Very interested in learning this approach!
For those interested in ES-RNN implement also check out N-BEATs: https://towardsdatascience.com/n-beats-beating-statistical-models-with-neural-nets-28a4ba4a4de8 https://github.com/philipperemy/n-beats/tree/master/examples
Currently working on a colab notebook. Hoping to extend the implementation to non-M4 data where you may have additional metadata (day of the week, holidays, etc.). Happy to share once it's done.
@samuelgneff How is your project going :)? Looking forward to your nb! Cheers
Please find a simple Colab NB example here: https://gist.github.com/florisrc/e9bc4cb054c6ad1c8f1ac43a9c21d09f
Please find a simple Colab NB example here: https://gist.github.com/florisrc/e9bc4cb054c6ad1c8f1ac43a9c21d09f
I added tensorflow 1 compatibility and a direct link to google colab here :+1: .
I'm excited to see the end-to-end example. I'd like to generalize this to non-M4 time series, and that'll be super helpful for it. Thanks!
Has anyone successfully generalized this to non-m4 time series and has a notebook of this? For example, the M3 time series?
