Supplementary Material for Optimal Transport Model Distributional Robustness Van-Anh Nguyen 1 Trung Le

This section presents all proofs in our work. It's worth noting that the experiments in Table 1 utilize an input resolution of 32x32, This noise can lead to a reduction in accuracy. WideResNet, to ensure the convergence of SGLD. This average prediction was obtained by aggregating the softmax predictions from all the base classifiers. Moreover, to ensure reliable uncertainty estimation, we employ calibrated uncertainty scores (Brier, NLL, ECE, and AAC).