Aug 2026· IEEE Transactions on Cybernetics· Vol PP· 0 citations
Medicine
TL;DR
A multicenter learning method that leverages the advantage of stochastic neural networks (SNNs) for feature uncertainty learning and induces multiple centers for each class of samples in latent space to fit data more delicately, named the multicenter SNN (MC-SNN).
Abstract
Enhancing the adversarial robustness of deep neural networks (DNNs) has become a prominent topic in the field of reliable artificial intelligence. Existing methods, no matter with standard training (ST) or adversarial training (AT), usually adopt a regular learning mode that induces a single center for each class of samples in the logit layer. However, due to the complex nature of data, a class of samples may have multiple dense regions, thereby following a mixed Gaussian distribution in latent space. The single-center learning mode makes it difficult for the network to fit data at a fine-grained level, thereby increasing the risks of some samples being located near the classification boundary, which degrades the model's adversarial robustness. In this article, we propose a multicenter learning method for robustness enhancement. It leverages the advantage of stochastic neural networks (SNNs) for feature uncertainty learning and induces multiple centers for each class of samples in latent space to fit data more delicately, named the multicenter SNN (MC-SNN). In addition, four AT-related strategies are introduced to propose MC-SNN-AT, in order to defend against a wider range of attacks. In a series of benchmark tests, both MC-SNN and MC-SNN-AT achieved state-of-the-art robustness. Furthermore, the training cost of MC-SNN is only about one-tenth of vanilla AT.
A method to analyze ANNs designed for image classification from an adversarial robustness perspective and implemented an ablation and fine-tuning strategy that successfully boosted the robustness of the ANNs against a variant of the Auto-PGD attack under different threat models.
Owing to the low occurrence frequencies of extreme financial risk events, risk prediction models are prone to being dominated by the majority of the normal samples encountered during the pretraining process when using financial risk data. This weakens the ability to effectively learn abnormal characteristics and thus reduces the sensitivity and stability of the developed predictive model. To overcome these limitations, a data class-balancing model, Financial Wasserstein Generative Adversarial Network with Gradient Penalty (FinWGAN-GP), is constructed on the basis of the Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) framework. Experimental results reveal that the FinWGAN-GP model achieves high-fidelity generation and expansion for high-risk samples, and a balanced dataset is conducive to improving the training effect of the early warning model. Through the integration of multisource heterogeneous data and Artificial Intelligence (AI) methods, the proposed model can identify potential financial risks earlier and more accurately, helping enhance the forward-looking and scientific nature of financial regulation.
Bo Xu, Lifei Lu, Yulin Lu et al.· Journal of Fintech and Busin...· 0 citations
Extensive experiments conducted on UCI and KEEL benchmark datasets demonstrate the superiority of the proposed IF-dRVFL and IF-edRVFL models over existing SOTA fuzzy and non-fuzzy approaches.
M. Sajid, A. Quadir, A. Rahaman et al.· 0 citations
This paper reveals that samples generated by a well-trained generative model are close to clean ones but far from adversarial ones, and proposes Consistency Model-based Adversarial Purification (CMAP), which optimizes vectors within the latent space of a pre-trained consistency model to generate samples for restoring clean data.
Shuhai Zhang, Jiahao Yang, Hui Luo et al.· IEEE Transactions on Pattern...· 0 citations
Deep learning models have achieved remarkable success across various domains, yet they remain vulnerable to adversarial examples, small carefully crafted perturbations of input images that cause models to make incorrect predictions. These adversarial examples are usually indistinguishable from the original input, yet the model classifies them incorrectly, which implies the lack of robustness of trained models. This work explores different neural network architectures, including fully connected networks, classical convolutional networks, and residual networks, under four types of adversarial attacks constrained by different L
p
norms. We evaluated attack success rates across multiple datasets and observed how different models behave when faced with various adversarial examples. All attacks are remarkably effective across all models and lead to misclassification almost every time. Next, we investigate how adversarial examples affect the internal representations of networks by analyzing the nearest neighbors and class manifold proximity across layers. Our results show that misclassification often occurs in the last couple of layers of the models, with variations depending on the dataset and the model used. In order to use a large model such as Residual Network 18 (ResNet-18), we apply principal component analysis to reduce unnecessary dimensions and to lower time complexity. We also analyzed how this reduction affects the results. This work highlights the importance of understanding not only if a model fails under a given attack but also how and where these failures occur within the network architecture.
Jana Poľašková, Iveta Bečková, Stefan Pócos et al.· PeerJ Computer Science· 0 citations
This work concerns adversarial training against the small-norm adversarial examples that arise from the inherent input instability of a trained deep neural network. Examples in this class are small as measured in the relative $\ell^2$-norm, and therefore lie in the neighborhood of the input on which the model acts approximately linearly, the regime in which the perturbation remains imperceptible. We first show that such examples can be computed directly from the trained network parameters, without input gradient iterations, by means of a linearization called the low-rank Householder expansion (LRHE). The expansion describes the composed affine map rather than any individual layer, and the directions it identifies are read from the activation pattern already available in the forward pass. We then propose a simple adversarial training scheme built on this construction. No differentiation with respect to the input is performed at any point: training requires only additional forward evaluations, with weight parameters updated by the standard backward pass, and the inner maximization of the usual min-max formulation is eliminated entirely. That such a regularizer exists is our main finding: the methods that dispense with the inner search all obtain their local geometry by differentiating with respect to the input, and we show this is not necessary. The regularizer costs the equivalent of $2.8$ PGD steps per epoch, an $8.7\times$ reduction relative to 40-step adversarial training on MNIST and below the cost of 3-step training. The resulting models match three-step PGD adversarial training for relative $\ell^2$ budgets $\varepsilon \le 0.02$ and 40-step training for $\varepsilon \le 0.012$, falling away beyond, consistent with the locality of the expansion.