贝叶斯神经网络与等变性结合，数据增强理论新突破，提升模型鲁棒性
arXiv:2606.26273v1 Announce Type: new Abstract: Symmetries are important for many deep learning tasks, ranging from applications in the sciences to me…
提出利用投影正则化实现近似等变性的新方法，为几何深度学习提供更灵活的理论框架
arXiv:2601.05028v2 Announce Type: replace Abstract: Equivariance is a powerful inductive bias in neural networks, improving generalisation and physica…
提出归一化等变性的结构先验，可应用于任意骨干网络的图像去噪，有效提升分布偏移健壮性。
arXiv:2605.08193v2 Announce Type: replace-cross Abstract: Normalization Equivariance (NE) is a structural prior that improves robustness to distributi…
用物理对称性破缺理论解释深层网络信息传播，提出Goldstone类似模式实现长距离相干传递，前沿交叉理论。
arXiv:2605.14685v1 Announce Type: cross Abstract: In physical systems, whenever a continuous symmetry is spontaneously broken, the system possesses ex…