Dynamic Routing Between Capsules

NeurIPS 2017 Sara Sabour • Nicholas Frosst • Geoffrey E Hinton

We use the length of the activity vector to represent the probability that the entity exists and its orientation to represent the instantiation parameters. Active capsules at one level make predictions, via transformation matrices, for the instantiation parameters of higher-level capsules. When multiple predictions agree, a higher level capsule becomes active.

Full paper


Task Dataset Model Metric name Metric value Global rank Compare
Image Classification MultiMNIST CapsNet Percentage error 5.2 # 1