Conformal quantile

[JudyJin]

trafficstars

Enhance the library with conformal quantile prediction torchts:

• Add quantile_err in loss.py
• Add calibration, calibration_pred, conformal_predict in model.py
• Modify fit, _step, validation_step in model.py

example:

• Add quantile-regression/conformal-quantile-regression_example.ipynb and uantile-regression/conformal-quantile-timeseries_example.ipynb by using the conformal quantile prediction

Potential problem:

1. Current model use validation_step to do the calibration adjustment, which may not be compatible if the user chooses to have the real validation set.
2. Problem in time-series prediction, if the training/test uses a different data window, the result may have lower accuracy.