Corrado Mio
Corrado Mio
I am using the latest version of sktime and the dump from github, and the previous post is based EXACTLY on the current implementation of NeuralForecastRNN and NeuralForecastLSTM. The current...
Currently I am integrating NeuralForecast & Darts in sktime, following the sktime phylosophy However, NeuralForecast is SLOW, VERY, VERY, VERY slow. The integration is PLAIN, just a SIMPLE wrapper to...
A sktime adapter to use NF is SIMPLE/stupid: it reorganizes y, X in the following way: 1. it converts sktime 'y' data into a dataframe with columns "unique_id" (an integer,...
[BUG] `predict` returns wrong index if inner mtypes are `numpy` - can we add custom output handling?
The example is attached in my first post
[BUG] `predict` returns wrong index if inner mtypes are `numpy` - can we add custom output handling?
I updated a little the code: I just added some prints. The idea of the method was to resolve the "index" problem. For example: I implemented a lot of time...
[BUG] `predict` returns wrong index if inner mtypes are `numpy` - can we add custom output handling?
There is not a "specific" use case. I am working on an article where I compare 1) classical TS models (ARIMA, GARCH, ...) vs tabular models applied to TS vs...