In this work, we study hallucinations in Neural Machine Translation (NMT), which lie at an extreme end on the spectrum of NMT pathologies. Firstly, we connect the phenomenon of …
Z Huang, G Niu, X Liu, W Ding… - Advances in Neural …, 2021 - proceedings.neurips.cc
Cross-modal matching, which aims to establish the correspondence between two different modalities, is fundamental to a variety of tasks such as cross-modal retrieval and vision-and …
In learning with noisy labels, the sample selection approach is very popular, which regards small-loss data as correctly labeled during training. However, losses are generated on-the …
Z Zhu, T Liu, Y Liu - … of the IEEE/CVF conference on …, 2021 - openaccess.thecvf.com
The presence of label noise often misleads the training of deep neural networks. Departing from the recent literature which largely assumes the label noise rate is only determined by …
X Li, T Liu, B Han, G Niu… - … conference on machine …, 2021 - proceedings.mlr.press
In label-noise learning, the transition matrix plays a key role in building statistically consistent classifiers. Existing consistent estimators for the transition matrix have been …
ZF Wu, T Wei, J Jiang, C Mao… - Proceedings of the …, 2021 - openaccess.thecvf.com
The existence of noisy data is prevalent in both the training and testing phases of machine learning systems, which inevitably leads to the degradation of model performance. There …
T Kim, J Ko, JH Choi, SY Yun - Advances in Neural …, 2021 - proceedings.neurips.cc
Modern deep neural networks (DNNs) become frail when the datasets contain noisy (incorrect) class labels. Robust techniques in the presence of noisy labels can be …
H Wei, L Tao, R Xie, B An - Advances in Neural Information …, 2021 - proceedings.neurips.cc
Learning with noisy labels is a practically challenging problem in weakly supervised learning. In the existing literature, open-set noises are always considered to be poisonous …
The recent breakthrough achieved by contrastive learning accelerates the pace for deploying unsupervised training on real-world data applications. However, unlabeled data …