CapsNet

CapsNet (Capsules Net) in Geoffrey E Hinton paper "Dynamic Routing Between Capsules"

What's New

What's New

Webinar

Pytorch Live Talks t.1 – Capsules and routing techniques (part 1/2), Alfredo Canziani 🆕

Implementations By Dataset

Toxic Comment Challenge (Kaggle)

chongjiujjin/capsule-net-with-gru

Abstract

We cover here the last and most interesting paper's abstract about Capsule Networks.

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

Excerpt from CAPSULES WITH INVERTED DOT-PRODUCT ATTENTION ROUTING, Yao-Hung Hubert Tsai, , Nitish Srivastava, Hanlin Goh, Ruslan Salakhutdinov, ICLR 2020 🆕

Documentation

Papers

CAPSULES WITH INVERTED DOT-PRODUCT ATTENTION ROUTING, Yao-Hung Hubert Tsai, , Nitish Srivastava, Hanlin Goh, Ruslan Salakhutdinov, ICLR 2020
Dynamic Routing Between Capsules, Sara Sabour, Nicholas Frosst, Geoffrey E Hinton, 26 Oct 2017
MATRIX CAPSULES WITH EM ROUTING, Anonymous authors, ICLR 2018
Does the Brain do Inverse Graphics?, Geoffrey Hinton, Alex Krizhevsky, Navdeep Jaitly, Tijmen Tieleman & Yichuan Tang, Department of Computer Science, University of Toronto, GSS 2012.
CapsuleGAN: Generative Adversarial Capsule Network Ayush Jaiswal, Wael AbdAlmageed, Premkumar Natarajan, USC Information Sciences Institute, Feb 2018
Brain Tumor Type Classification via Capsule Networks Parnian Afshar, Arash Mohammadi, Konstantinos N. Plataniotis
Improved Explainability of Capsule Networks: Relevance Path by Agreement Atefeh Shahroudnejad, Arash Mohammadi, Konstantinos N. Plataniotis
Investigating Capsule Networks with Dynamic Routing for Text Classification Wei Zhao, Jianbo Ye, Min Yang, Zeyang Lei, Suofei Zhang, Zhou Zhao
Capsules for Object Segmentation Rodney LaLonde, Ulas Bagci
Investigating Capsule Networks with Dynamic Routing for Text Classification Wei Zhao, Jianbo Ye3, Min Yang1, Zeyang Lei4 , Soufei Zhang5 , Zhou Zhao6
The Multi-Lane Capsule Network (MLCN) Vanderson M. do Rosario, Edson Borin, Mauricio Bretenitz Jr., Feb 2019
From Attention in Transformers to Dynamic Routing in Capsule Nets
TextCaps : Handwritten Character Recognition with Very Small Datasets
DeepCaps: Going Deeper with Capsule Networks
Avoiding Implementation Pitfalls of "Matrix Capsules with EM Routing" by Hinton et al. Ashley Gritzman, Aug 2019
A Neural-Symbolic Architecture for Inverse Graphics Improved by Lifelong Meta-Learning
Adding Intuitive Physics to Neural-Symbolic Capsules Using Interaction Networks🆕
Capsule Routing via Variational Bayes
Building Deep, Equivariant Capsule Networks
Stacked Capsule Autoencoders, Adam R. Kosiorek, Sara Sabour, Yee Whye Teh, Geoffrey E. Hinton, 17 Jun 2019, revised 2 Dec 2019
Capsule Routing via Variational Bayes, Fabio De Sousa Ribeiro, Georgios Leontidis, Stefanos Kollias, 27 May 2019 (v1), last revised 3 Dec 2019
CAPSULES WITH INVERTED DOT-PRODUCT ATTENTION ROUTING, Yao-Hung Hubert Tsai, , Nitish Srivastava, Hanlin Goh, Ruslan Salakhutdinov, ICLR 2020 🆕

Articles

Tutorials

Presentations

Dynamic Routing Between Capsules

Webinar 🆕

Discussion Groups

Official Implementations

The implementations has been considered to be official since the authors were directly involved in the papers as co-authors or they had some references.