Deep4Cast Documentation

Decision making incorporating uncertainty.

This package is under active development. Things may change :-).

Deep4Cast is a scalable machine learning package implemented in Python and Torch. It has a front-end API similar to scikit-learn. It is designed for medium to large time series data sets and allows for modeling of forecast uncertainties.

The network architecture is based on WaveNet. Regularization and approximate sampling from posterior predictive distributions of forecasts are achieved via Concrete Dropout.

Examples

Tutorial: Forecasting Github Daily Active Users

Authors

• Toby Bischoff
• Austin Gross
• Kenneth Tran

References

• Concrete Dropout is used for approximate posterior Bayesian inference.
• Wavenet is used as encoder network.