Learning with Symmetric Label Noise: The Importance of Being Unhinged

Rooyen, Brendan van; Menon, Aditya Krishna; Williamson, Robert C.

doi:10.48550/arxiv.1505.07634

Cited by 7 publications

(7 citation statements)

References 9 publications

(15 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following [14,16], we shuffle the labels of the training set by a noise transition matrix Q, where Q i j = Pr[ ỹ = j|y = i] denotes the probability of flipping class i to j. The widely-used structures of Q, i.e., symmetry flipping [15,23] and pair flipping [3], are adopted in our study. Note that, consistent to [3,15,23], we validate our NIB module with different noise ratios, denoting as 'symmetry-10%', 'symmetry-20%', 'symmetry-40%' and 'pair-10%'.…”

Section: Experimental Settingsmentioning

confidence: 99%

“…The widely-used structures of Q, i.e., symmetry flipping [15,23] and pair flipping [3], are adopted in our study. Note that, consistent to [3,15,23], we validate our NIB module with different noise ratios, denoting as 'symmetry-10%', 'symmetry-20%', 'symmetry-40%' and 'pair-10%'. For example, the 'symmetry-10%' represents that 10% of the labels have been symmetrically flipped to be noisy labels.…”

Section: Experimental Settingsmentioning

confidence: 99%

See 1 more Smart Citation

Alleviating Noisy-label Effects in Image Classification via Probability Transition Matrix

Zhang¹,

Li²,

Wei³

et al. 2021

Preprint

View full text Add to dashboard Cite

Deep-learning-based image classification frameworks often suffer from the noisy label problem caused by the inter-observer variation. Recent studies employed learningto-learn paradigms (e.g., Co-teaching and JoCoR) to filter the samples with noisy labels from the training set. However, most of them use a simple cross-entropy loss as the criterion for noisy label identification. The hard samples, which are beneficial for classifier learning, are often mistakenly treated as noises in such a setting, since both the hard samples and the ones with noisy labels lead to a relatively larger loss value than the easy cases. In this paper, we propose a plugin module, namely noise ignoring block (NIB), consisting of a probability transition matrix and an inter-class correlation (IC) loss, to separate the hard samples from the mislabeled ones, and further boost the accuracy of image classification network trained with noisy labels. Concretely, our IC loss is calculated as Kullback-Leibler divergence between the network prediction and the accumulative soft label generated by the probability transition matrix. Such that, with lower value of IC loss, the hard cases can be easily distinguished from mislabeled cases. Extensive experiments are conducted on natural and medical image datasets (CIFAR-10 and ISIC 2019). The experimental results show that our NIB module consistently improves the performances of the state-of-the-art robust training methods. IntroductionWitnessing the success of deep neural networks (DNNs) for computer vision tasks [12,21], an increasing number of researchers began to implement deep-learning-based approaches

show abstract

Section: Experimental Settingsmentioning

confidence: 99%

Section: Experimental Settingsmentioning

confidence: 99%

Alleviating Noisy-label Effects in Image Classification via Probability Transition Matrix

Zhang¹,

Li²,

Wei³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…In this way, adapting the meta-model on the support set causes inaccurate task-specific parameters and leads to negative impacts on the meta-training process. Specifically, for miniImagenet, we apply the symmetry flipping on the labels of the support set [27]. The default ratio of noisy tasks is set as 0.6.…”

Section: Meta-learning With Noisementioning

confidence: 99%

Meta-learning with an Adaptive Task Scheduler

Yao¹,

Wang²,

Wei³

et al. 2021

Preprint

View full text Add to dashboard Cite

To benefit the learning of a new task, meta-learning has been proposed to transfer a well-generalized meta-model learned from various meta-training tasks. Existing meta-learning algorithms randomly sample meta-training tasks with a uniform probability, under the assumption that tasks are of equal importance. However, it is likely that tasks are detrimental with noise or imbalanced given a limited number of meta-training tasks. To prevent the meta-model from being corrupted by such detrimental tasks or dominated by tasks in the majority, in this paper, we propose an adaptive task scheduler (ATS) for the meta-training process. In ATS, for the first time, we design a neural scheduler to decide which meta-training tasks to use next by predicting the probability being sampled for each candidate task, and train the scheduler to optimize the generalization capacity of the metamodel to unseen tasks. We identify two meta-model-related factors as the input of the neural scheduler, which characterize the difficulty of a candidate task to the meta-model. Theoretically, we show that a scheduler taking the two factors into account improves the meta-training loss and also the optimization landscape. Under the setting of meta-learning with noise and limited budgets, ATS improves the performance on both miniImageNet and a real-world drug discovery benchmark by up to 13% and 18%, respectively, compared to state-of-the-art task schedulers.Recently, a few studies have started to consider introducing a task scheduler into meta-learning, by adjusting the class selection strategy for construction of each few-shot classification task [17,19],

show abstract

“…Much effort has also been devoted to the sampling algorithms to overcome the data imbalance, such as SMOTE [56], and ADASYN [57]. As for the relabeling method, [58], [59] uniformly change the label to other classes with a constant flip probability, and the trained models appear to be more resistant to the label noise. In contrast, our proposed method adaptive refined labeling is designed to adaptively refine the similar samples with opposite labels to boost the performance of predictors in the quantitative trading scenarios.…”

Section: Related Workmentioning

confidence: 99%