Machine Intelligence (MI) Rider Task 4: Dynamic Routing Between Capsules Implementation

reportActive / Technical Report | Accesssion Number: AD1079647 | Open PDF

Abstract:

This paper implemented Geoffrey E Hinton proposed implementation of Dynamic Routing between Capsules. The intent was to confirm the claims asserted by the author about the ability of Capsnet being rotational invariant to changes in the position of an image. Some claims validated to be true; however, other claims did not live to all the hype created. Results were collected by running extensive trials of Capsnet with different datasets such as the MNIST and Cifar10 datasets. Suggestion and directions are provided within the paper on future actions on how to move forward with this technology potential.

Author(s):
Franco, Milvio ; Khan, Simon
Author Organization(s):
Air Force Research Laboratory, Information Directorate
Funding Organization(s):
Publication Date:
2019 Aug 01
Pagination:
0

Security Markings

RECORD

Collection: TRECMS
Subject Terms
Modernization Areas:
AI and Machine Learning
Communities of Interest:
Energy and Power Technologies
Descriptor(s):
neural networks, air force, information systems, artificial intelligence, convolutional neural networks, artificial intelligence software, artificial intelligence computing, computer languages, probability, government procurement, governments, contracts, information exchange, accuracy, algorithms, application software, computer programming
Keyword(s):
Convolutional Neural Network, Dynamic capsules, non-linearity functions, machine intelligence
Subject Categories:
Mathematical and Computer Sciences
Update Date:
2019 Dec 04