CODEs: Chamfer Out-of-Distribution Examples against Overconfidence Issue

Tang, Keke; Miao, Dingruibo; Peng, Weilong; Wu, Jianpeng; Shi, Yawen; Gu, Zhaoquan; Tian, Zhihong; Wang, Wenping

doi:10.1109/iccv48922.2021.00119

Cited by 16 publications

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As shown in the simulation, the designed self-triggered algorithm is reasonable in terms of avoiding Zeno behavior. However, just quasisynchronization and quasi-anti-synchronization are considered, from the projective factor point of view, the time-varying factor is more universal for application, especially in social networks (Cheng et al, 2022), object/scene reconstruction (Tang et al, 2021), and 3D object recognition etc. Therefore, in future study, the obtained scheme will be developed for projective quasi-synchronization with time-varying projective factors and more practical networks will be considered.…”

Section: Discussionmentioning

confidence: 99%

Projective quasi-synchronization of coupled memristive neural networks with uncertainties and impulsive effect

Yuan

Luo

et al. 2022

Front. Neurorobot.

View full text Add to dashboard Cite

The dynamic behavior of memristive neural networks (MNNs), including synchronization, effectively keeps the robotic stability against numerous uncertainties from the mimic of the human brain. However, it is challenging to perform projective quasi-synchronization of coupled MNNs with low-consumer control devices. This is partly because complete synchronization is difficult to realize under various projective factors and parameter mismatch. This article aims to investigate projective quasi-synchronization from the perspective of the controller. Here, two approaches are considered to find the event-triggered scheme for lag synchronization of coupled MNNs. In the first approach, the projective quasi-synchronization issue is formulated for coupled MNNs for the first time, where the networks are combined with time-varying delays and uncertainties under the constraints imposed by the frequency of controller updates within limited system communication resources. It is shown that our methods can avoid the Zeno-behavior under the newly determined triggered functions. In the second approach, following classical methods, a novel projective quasi-synchronization criterion that combines the nonlinear property of the memristor and the framework of Lyapunov-Krasovskii functional (LKF) is proposed. Simulation results indicate that the proposed two approaches are useful for coupled MNNs, and they have less control cost for different types of quasi-synchronization.

show abstract

Section: Discussionmentioning

confidence: 99%

Projective quasi-synchronization of coupled memristive neural networks with uncertainties and impulsive effect

Yuan

Luo

et al. 2022

Front. Neurorobot.

View full text Add to dashboard Cite

show abstract

“…The most commonly used training criterion in training deep neural networks is the softmax cross-entropy loss. However, previous works (Tang et al 2021;Liu et al 2020) show that directly training with this loss results in overconfidence issues, where the maximum softmax activation value always approaches one in despite of whether the data is from training data distribution or not. Previous works have shown that other criteria such as the Helmholtz free energy (Liu et al 2020) or the maximum logit value (Hendrycks et al 2019) are better confidence scores than the maximum softmax value.…”

Section: Related Workmentioning

confidence: 97%

“…Rehearsal-based methods tackle catastrophic forgetting either by keeping a small set of old training examples in memory (Tao et al 2020a;Dong et al 2021;Liu et al 2022;Yang et al 2022a,b) or using synthesized data produced by generative models (Shin et al 2017). By using the rehearsal buffer for knowledge distillation and regularization, rehearsal-based methods have achieved state-of-theart results on various benchmarks (Douillard et al 2022;Joseph et al 2022;Zhang et al 2022).…”

Section: Related Workmentioning

confidence: 99%

Isolation and Impartial Aggregation: A Paradigm of Incremental Learning without Interference

Wang

Huang

et al. 2023

AAAI

View full text Add to dashboard Cite

This paper focuses on the prevalent stage interference and stage performance imbalance of incremental learning. To avoid obvious stage learning bottlenecks, we propose a new incremental learning framework, which leverages a series of stage-isolated classifiers to perform the learning task at each stage, without interference from others. To be concrete, to aggregate multiple stage classifiers as a uniform one impartially, we first introduce a temperature-controlled energy metric for indicating the confidence score levels of the stage classifiers. We then propose an anchor-based energy self-normalization strategy to ensure the stage classifiers work at the same energy level. Finally, we design a voting-based inference augmentation strategy for robust inference. The proposed method is rehearsal-free and can work for almost all incremental learning scenarios. We evaluate the proposed method on four large datasets. Extensive results demonstrate the superiority of the proposed method in setting up new state-of-the-art overall performance. Code is available at https://github.com/iamwangyabin/ESN.

show abstract

“…Self-awareness: Self-aware systems have some ability to measure their limitations or predict failures. This includes out-of-distribution detection [78], [79], [80], [81] or open set recognition [82], [83], [84], [85], where classifiers are trained to reject non-sensical images, adversarial attacks, or images from classes on which they were not trained. All these problems require the classifier to produce a confidence score for image rejection.…”

Section: Discriminant Explanation Thresholdmentioning

confidence: 99%

A Generalized Explanation Framework for Visualization of Deep Learning Model Predictions

Wang

Vasconcelos

2023

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

Attribution-based explanations are popular in computer vision but of limited use for fine-grained classification problems typical of expert domains, where classes differ by subtle details. In these domains, users also seek understanding of "why" a class was chosen and "why not" an alternative class. A new GenerAlized expLanatiOn fRamEwork (GALORE) is proposed to satisfy all these requirements, by unifying attributive explanations with explanations of two other types. The first is a new class of explanations, denoted deliberative, proposed to address the "why" question, by exposing the network insecurities about a prediction. The second is the class of counterfactual explanations, which have been shown to address the "why not" question but are now more efficiently computed. GALORE unifies these explanations by defining them as combinations of attribution maps with respect to various classifier predictions and a confidence score. An evaluation protocol that leverages object recognition (CUB200) and scene classification (ADE20K) datasets combining part and attribute annotations is also proposed. Experiments show that confidence scores can improve explanation accuracy, deliberative explanations provide insight into the network deliberation process, the latter correlates with that performed by humans, and counterfactual explanations enhance the performance of human students in machine teaching experiments.

show abstract

CODEs: Chamfer Out-of-Distribution Examples against Overconfidence Issue

Cited by 16 publications

References 21 publications

Projective quasi-synchronization of coupled memristive neural networks with uncertainties and impulsive effect

Projective quasi-synchronization of coupled memristive neural networks with uncertainties and impulsive effect

Isolation and Impartial Aggregation: A Paradigm of Incremental Learning without Interference

A Generalized Explanation Framework for Visualization of Deep Learning Model Predictions

Contact Info

Product

Resources

About