Capsules with Inverted Dot-Product Attention Routing
This work addresses the challenge of making capsule networks more efficient and effective for complex real-world tasks, representing an incremental improvement over existing routing methods.
The paper tackles the problem of improving routing algorithms in capsule networks by introducing a new method based on inverted dot-product attention, which enhances performance on benchmark datasets like CIFAR-10 and CIFAR-100, achieving results comparable to ResNet-18 with 4x fewer parameters.
We introduce a new routing algorithm for capsule networks, in which a child capsule is routed to a parent based only on agreement between the parent's state and the child's vote. The new mechanism 1) designs routing via inverted dot-product attention; 2) imposes Layer Normalization as normalization; and 3) replaces sequential iterative routing with concurrent iterative routing. When compared to previously proposed routing algorithms, our method improves performance on benchmark datasets such as CIFAR-10 and CIFAR-100, and it performs at-par with a powerful CNN (ResNet-18) with 4x fewer parameters. On a different task of recognizing digits from overlayed digit images, the proposed capsule model performs favorably against CNNs given the same number of layers and neurons per layer. We believe that our work raises the possibility of applying capsule networks to complex real-world tasks. Our code is publicly available at: https://github.com/apple/ml-capsules-inverted-attention-routing An alternative implementation is available at: https://github.com/yaohungt/Capsules-Inverted-Attention-Routing/blob/master/README.md