Robust Calibration For Improved Weather Prediction Under Distributional Shift

In this paper, we present results on improving out-of-domain weather prediction and uncertainty estimation as part of the $\mathtt{\text{Shifts}}\mathtt{\text{Challenge}}\mathtt{\text{on}}\mathtt{\text{Robustness}}\mathtt{\text{and}}\mathtt{\text{Uncertainty}}\mathtt{\text{under}}\mathtt{\text{Real-World}}\mathtt{\text{Distributional}}\mathtt{\text{Shift}}$ challenge. We find that by leveraging a mixture of experts in conjunction with an advanced data augmentation technique borrowed from the computer vision domain, in conjunction with robust $\mathit{\text{post-hoc}}$ calibration of predictive uncertainties, we can potentially achieve more accurate and better-calibrated results with deep neural networks than with boosted tree models for tabular data. We quantify our predictions using several metrics and propose several future lines of inquiry and experimentation to boost performance.