[FEAT] Implicit Quantile Networks

[elephaint]

Features:

• Adds Implicit Quantile Loss as a loss function to facilitate Implicite Quantile Networks for all architectures. This implementation is based on the GluonTS implementation, following the original Gouttes IQN paper.
• Cleanup moving parts to base_model out of the individual architectures, removing exceptions out of individual architectures.
• Adds a number of class attributes to models that indicate whether the model supports exogenous.
• Fixes some architecture errors, mostly in *former models.

Tests:

• Test for a range of architectures on AirPassengers

References:

• IQN
• IQN for forecasting

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB