Abstract:Capsule networks (CapsNet) are recently proposed neural network models containing newly introduced processing layer, which are specialized in entity representation and discovery in images. CapsNet is motivated by a view of parse tree-like information processing mechanism and employs an iterative routing operation dynamically determining connections between layers composed of capsule units, in which the information ascends through different levels of interpretations, from raw sensory observation to semantically meaningful entities represented by active capsules. The CapsNet architecture is plausible and has been proven to be effective in some image data processing tasks, the newly introduced routing operation is mainly required for determining the capsules’ activation status during the forward pass. However, its influence on model fitting and the resulted representation is barely understood. In this work, we investigate the following: 1) how the routing affects the CapsNet model fitting; 2) how the representation using capsules helps discover global structures in data distribution, and; 3) how the learned data representation adapts and generalizes to new tasks. Our investigation yielded the results some of which have been mentioned in the original paper of CapsNet, they are: 1) the routing operation determines the certainty with which a layer of capsules pass information to the layer above and the appropriate level of certainty is related to the model fitness; 2) in a designed experiment using data with a known 2D structure, capsule representations enable a more meaningful 2D manifold embedding than neurons do in a standard convolutional neural network (CNN), and; 3) compared with neurons of the standard CNN, capsules of successive layers are less coupled and more adaptive to new data distribution.

A Novel Capsule Network Based on Multi-Order Descartes Extension Transformation

Deep multi-prototype capsule networks

DE-CapsNet: A Diverse Enhanced Capsule Network with Disperse Dynamic Routing

A novel capsule network based on deep routing and residual learning

A Context-aware Capsule Network for Multi-label Classification

A novel dense capsule network based on dense capsule layers

Multi-Lane Capsule Network for Classifying Images With Complex Background

DeepCaps: Going Deeper With Capsule Networks

Cv-CapsNet: Complex-Valued Capsule Network

Capsule Network Performance on Complex Data

Capsule Networks With Residual Pose Routing

A Multi-prototype Capsule Network for Image Recognition with High Intra-class Variations

On Learning and Learned Data Representation by Capsule Networks

Ddrm-Capsnet: Capsule Network Based On Deep Dynamic Routing Mechanism For Complex Data

RS-CapsNet: an Advanced Capsule Network.

Efficient-CapsNet: capsule network with self-attention routing

A lightweight capsule network via channel-space decoupling and self-attention routing

Path Capsule Networks

Training Deep Capsule Networks with Residual Connections

The Analysis Between Traditional Convolution Neural Network and CapsuleNet

Modified Capsule Network For Object Classification