Jan 26, 2022 · Experiments on CIFAR-10, CIFAR-100, and SVHN show that regularizing diversity can have a significant impact on calibration and robustness, as ...

Sep 20, 2022 · We introduce a new diversity regularizer for classification tasks that uses out-of-distribution samples and increases the overall accuracy, calibration and out ...

Sep 27, 2022 · We introduce a new diversity regularizer for classification tasks that uses out-of-distribution samples and increases the overall accuracy, ...

Jan 26, 2022 · Experiments on CIFAR-10, CIFAR-100, and SVHN show that regularizing diversity can have a significant impact on calibration and robustness, as ...

Jan 26, 2022 · Calibration and uncertainty estimation are crucial top- ics in high-risk environments. We introduce a new diversity regularizer for ...

Improving Robustness and Calibration in Ensembles with Diversity Regularization ... Improving adversarial robustness of ensembles with diversity training (2019) ...

Relying on ensemble diversity strategies to improve adversarial robustness has been investi- gated in several papers, but the gains provided by ...

This work systematically evaluates the viability of explicitly regularizing ensemble diversity to improve calibration on in-distribution data as well as ...

The contributions of the present work are the followings: 1) A novel and simple anti-regularization strategy is proposed to increase deep ensemble diversity. 2) ...

The AUGMIX method (Hendrycks et al., 2020) aims to make models robust to out-of-distribution data by expos- ing the model to a wide variety of augmented images.