https://github.com/nixtla logo
#general
Title
# general
m

Max (Nixtla)

08/17/2022, 2:48 PM
@Valeriy: its settled, NeuralProphet is also sort of broken.
πŸ”₯ 1
k

Kin Gtz. Olivares

08/17/2022, 2:58 PM
Guys at Facebook should stop burning millions in the Metaverse and build reasonable forecasting algorithms.
πŸ™Œ 1
πŸ”₯ 1
v

Valeriy

08/17/2022, 2:59 PM
Haha you guys have smokedπŸ”₯ Neural Prophet. Great work from Nixtla again
As everyone knows I am a huge sceptic of Prophet, I have done some initial limited explorations of Neural Prophet beginning of this year but then didn't have time as started new role. https://valeman.medium.com/benchmarking-facebook-prophet-53273c3ee9c6 https://valeman.medium.com/benchmarking-neural-prophet-part-i-neural-prophet-vs-prophet-252990763468 https://valeman.medium.com/benchmarking-neural-prophet-part-ii-exploring-electricity-dataset-4cade7b2431c TL;DR Results were somewhat better than Prophet as including autoregressive terms helped, but not conclusive. Great comprehensive work evaluating it across multiple datasets.
@Max (Nixtla) it is settled bigly, I didn't pin big hopes on it as it is still curve fitter but AR terms were improvement even then my own tests indicated β€œdeepAR” didn't improve much compared to linear AR. Even then to see it losing to ES is big news.
@Max (Nixtla) @Kin Gtz. Olivares quick question - what sort of ES model is under the hood? Is there some document describing it, presumably it is some sort of similar to auto ARIMA with automatic hyperparameter optimisation?
m

Max (Nixtla)

08/21/2022, 3:27 PM
Hi @Valeriy, we have been working on the documentation of the library. In the model section of the library, you can find some further info https://nixtla.github.io/statsforecast/models.html#exponential-smoothing-methods Basically, we ported the ETS from Rob. (https://otexts.com/fpp3/expsmooth.html) here is more info on the optimization algorithm itself: https://otexts.com/fpp3/expsmooth.html If you miss anything in the docs, please tag @MarianaR so we can include it and make the docs more straightforward.
πŸ‘ 1
I was editing the post and I erased the other thing I wrote about the nelder mead. Either way, we are preparing a post explaining that particular aspect and making that particular optimization method available to the community.
v

Valeriy

08/21/2022, 3:41 PM
Nice, thanks for sharing. Got to see SKTime opened their lib to other implementations.
k

Kin Gtz. Olivares

08/21/2022, 3:42 PM
They had to Theirs/StatsModels is broken
5 Views