[WIP] Add an embedding layer for the months

ECMWFCode4Earth / ml_drought

Machine learning to better predict and understand drought. Moving github.com/ml-clim

https://ml-clim.github.io/drought-prediction/

90 stars 18 forks source link

[WIP] Add an embedding layer for the months #120

Open gabrieltseng opened 5 years ago

gabrieltseng commented 5 years ago

As per Yasser's comments, replaces the one hot encoding with an embedding.

~Currently, the size of the embedding is fixed at 12, but if this works well this could become an additional argument.~ The size of the embedding is an additional argument - if None is passed, the prediction month isn't passed to the model.

Also, removes the EALSTM cell written using nn.Modules, since the original one works fine with shap.

tommylees112 commented 4 years ago

@gabrieltseng is this still good to merge?