用信息瓶颈原理改进胶囊网络，抗噪更强且速度更快。
IBCapsNet: Information Bottleneck Capsule Network for Noise-Robust Representation Learning
- 用单次变分聚合替代迭代路由，降低计算开销。
- 在噪声数据上平均提升17.10%（加性噪声），精度达92.01%（SVHN）。
- 训练快2.54倍，推理吞吐高3.64倍，参数减少4.66%。
胶囊网络（CapsNets）在建模层级空间关系上表现优异，但存在两大局限：因迭代动态路由导致计算成本高，且对输入噪声鲁棒性差。为此，我们提出基于信息瓶颈（IB）原理的IBCapsNet新架构。不同于迭代路由，IBCapsNet采用单次变分聚合机制：先将初级胶囊压缩为全局上下文表示，再通过类别特定的变分自编码器（VAEs）推断受KL散度正则化的潜在胶囊。该设计实现高效推理的同时，可内在过滤噪声。在MNIST、Fashion-MNIST、SVHN和CIFAR-10上的实验表明，IBCapsNet在无噪声数据下准确率与CapsNet相当（如MNIST达99.41%，SVHN达92.01%），但在四类合成噪声下显著更优——对钳制加性噪声和乘性噪声的平均提升分别为+17.10%和+14.54%。此外，其训练速度提升2.54倍，推理吞吐量提高3.64倍，模型参数减少4.66%。本工作将信息论表征学习与胶囊网络融合，为构建稳健、高效、可解释的深度模型提供了原则性路径。代码已开源：https://github.com/cxiang26/IBCapsnet。
原文摘要 · Abstract (English)
Capsule networks (CapsNets) are superior at modeling hierarchical spatial relationships but suffer from two critical limitations: high computational cost due to iterative dynamic routing and poor robustness under input corruptions. To address these issues, we propose IBCapsNet, a novel capsule architecture grounded in the Information Bottleneck (IB) principle. Instead of iterative routing, IBCapsNet employs a one-pass variational aggregation mechanism, where primary capsules are first compressed into a global context representation and then processed by class-specific variational autoencoders (VAEs) to infer latent capsules regularized by the KL divergence. This design enables efficient inference while inherently filtering out noise. Experiments on MNIST, Fashion-MNIST, SVHN and CIFAR-10 show that IBCapsNet matches CapsNet in clean-data accuracy (achieving 99.41% on MNIST and 92.01% on SVHN), yet significantly outperforms it under four types of synthetic noise - demonstrating average improvements of +17.10% and +14.54% for clamped additive and multiplicative noise, respectively. Moreover, IBCapsNet achieves 2.54x faster training and 3.64x higher inference throughput compared to CapsNet, while reducing model parameters by 4.66%. Our work bridges information-theoretic representation learning with capsule networks, offering a principled path toward robust, efficient, and interpretable deep models. Code is available at https://github.com/cxiang26/IBCapsnet
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。