search

JoaquinAmatRodrigo / skforecast

Time series forecasting with machine learning models
https://skforecast.org
BSD 3-Clause "New" or "Revised" License
1.07k stars 123 forks source link

Deep learning RNN - regularizers #757

Closed AVPokrovsky closed 1 month ago

AVPokrovsky commented 1 month ago

Hi, can I set up Layer weight regularizers?

https://keras.io/2.15/api/layers/regularizers/

JoaquinAmatRodrigo commented 1 month ago

@FernandoCarazoMelo

FernandoCarazoMelo commented 1 month ago

Hey @AVPokrovsky,

Thank you for using skforecast and for your insightful comment!

Currently, the create_and_compile utility function does not support regularizers. The function includes the following layers: input > RNN_1 > RNN_2 > RNN_n > dense.

However, you can create any TensorFlow model and pass it to the forecasterRNN. In that model, you can use any regularizer or other RNN layers. You can easily follow the function template.

Also, if you have any ideas to improve the utility function, please feel free to contribute to the library!

Let me know if you have any other questions or suggestions.

FernandoCarazoMelo commented 1 month ago

@JoaquinAmatRodrigo , I believe this is already answered. This is not planned to be included in the short term.