search

cchallu / n-hits

170 stars 25 forks source link

Question on n_time_in #9

Closed aminshabani closed 2 years ago

aminshabani commented 2 years ago

Hi,

Thank you for publishing your code and also thanks for your interesting paper. I am now trying to use your code but I am not sure if I need to update the hyper opt space for n_time_in?

The current settings in nhits_multivariate is set to 'n_time_in': hp.choice('n_time_in', [5*args.horizon]) which results 960 inputs for a horizon length of 192, I was wondering is it the one used in your experiments or should I changed it to 96?

Thanks

cchallu commented 2 years ago

Hi @aminshabani, thanks for trying our model.

To replicate the paper's results, you should use the default hyperparameter grid search; we use the grid in nhits_multivariate for all benchmark datasets and horizons.

However, the optimal configuration might be different for other datasets and tasks. We have a forecasting general-purpose library with the N-HiTS model in https://github.com/Nixtla/neuralforecast.