A Review Paper On Understanding Capsule Networks

  Prasham Lalit Shah,  Tanay Krishnakumar Gupta,  Jash Bhupendra Dhakad,  Mitchell Raymond D’silva

A CapsNet consists of capsules instead of neurons. A capsule is a group of neurons whose output predicts different features of the same object. Every layer of a capsule neural network consists of many capsules [1]. Capsule Neural Network or CapsNet is a type of artificial neural network (ANN) used in machine learning system that can be used to better represent hierarchal relationships. The network has been developed to overcome the limitations associated with convolutional neural networks (CNN) such as back-propagation, translation invariance and pooling layers by adding capsules to a CNN and reusing the output from several capsules to form more stable representations. The concept of CapsNet can be extended to various applications like Dynamic Routing, EM Routing and other image classification problems. This paper also states the characteristics, advantages and disadvantages of using capsule networks. It also tells us about the future scope of using CapsNet technology.

Keywords- Capsule Networks, Convolutional Neural Networks, Artificial Neural Network.

Unique Identification Number - IJEDR1804013
Page Number(s) - 58-65
Pubished in - Volume 6 | Issue 4 | October 2018
DOI (Digital Object Identifier) -   
Publisher - IJEDR (ISSN - 2321-9939)

  Prasham Lalit Shah,  Tanay Krishnakumar Gupta,  Jash Bhupendra Dhakad,  Mitchell Raymond D’silva,   "A Review Paper On Understanding Capsule Networks", International Journal of Engineering Development and Research (IJEDR), ISSN:2321-9939, Volume.6, Issue 4, pp.58-65, October 2018, Available at :http://www.ijedr.org/papers/IJEDR1804013.pdf