近日，本课题组在人工智能领域取得新进展，2篇研究论文分别被KDD2026和IJCAI2026录用。ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) 是数据挖掘与数据科学领域历史悠久的国际学术会议之一，被中国计算机学会(CCF)列为A类推荐会议。International Joint Conference on Artificial Intelligence (IJCAI) 是全球人工智能领域最具影响力的综合性顶级学术会议之一。

硕士生王嘉仪为第一作者的论文《Sparse Additive Models for Domain Generalization》针对域泛化问题中高维数据及黑盒网络架构透明度不足所带来的挑战，将稀疏可加模型引入到域泛化框架中，提出了两种实现方式，旨在实现结构化特征选择并提升可解释性。该文从理论上推导了两种实现方式的泛化误差界，并证明了特征选择的一致性，最后通过实验验证了该方法在处理高维数据时的有效性与鲁棒性。该工作是课题组前期可加模型（NeurIPS’17, ICML’20, NeurIPS’20, ICLR’22, ICML’23）研究的自然拓展。

【英文摘要】Vision Transformers (ViTs) exhibit notable susceptibility to adversarial attacks, presenting a significant challenge for their deployment in security-sensitive applications. Despite their considerable empirical successes, a rigorous theoretical foundation for ViT's adversarial generalization behavior has not been adequately established. To address this limitation, we leverage empirical Rademacher complexity to analyze the mechanism of perturbation accumulation through deep ViTs layers. We establish a high-probability generalization bound for ViTs in classification tasks under adversarial settings. Our theoretical framework elucidates the roles of several factors in mitigating perturbation effects, norm regularization of weight matrices (in both MLP and attention modules) and depth-wise propagation constraints on layer-wise norms. Extensive experiments on benchmark datasets corroborate our theoretical insights, bridging the gap between ViTs architecture design and adversarial robustness.

硕士生江子文为第一作者的论文《Generalization Analysis for Adversarial Vision Transformers》基于经验 Rademacher 复杂度，深刻揭示了 ViT 在对抗扰动下的泛化机理。针对非凸优化难题，该研究通过构建代理损失上界，推导出了紧致的高概率泛化界，并明确了泛化误差与特征维度的对数依赖关系。理论指出，通过权重范数约束与注意力稀疏性可有效抑制深层网络的扰动累积。多数据集实验充分证实了相关发现对缩小泛化的有效性。该工作为构建高可靠的鲁棒视觉模型提供了理论支撑，是课题组对抗泛化系列研究（IJCAI’23/24, ICLR’25, ICML’25）的进一步延伸。

【英文摘要】Machine learning models continue to face challenges in out-of-distribution (OOD) generalization, where Domain Generalization (DG) aims to improve performance on unseen domains under distributional shifts. A prevalent paradigm in DG focuses on learning domain-invariant feature representations. However, the feature representations learned by such current methods exhibit weak interpretability. To bridge this gap, we propose Sparse Additive Domain Generalization (SpADG). Specifically, we incorporate an additive structure into the DG framework and employ lq,1-norm regularization to induce sparsity, thereby enabling structured feature selection and enhancing interpretability. We present two distinct realizations: an additive kernel-based formulation and a neural additive model-based approach. The former leverages the representer theorem for flexible data adaptation, while the latter can learn highly nonlinear shape functions. Theoretically, we derive generalization error bounds for both realizations and prove the feature selection consistency of our method under rate-scaled regularization condition. Empirical evaluations on synthetic and real-world datasets validate the effectiveness of SpADG, particularly its robustness in high-dimensional settings.

上述论文通讯作者分别为信息学院李函副教授。