Kinch, Martin W. ORCID: https://orcid.org/0000-0002-4737-7118, Melis, Wim J.C. ORCID: https://orcid.org/0000-0003-3779-8629 and Keates, Simeon ORCID: https://orcid.org/0000-0002-2826-672X (2017) Reviewing the current state of machine learning for artificial intelligence with regards to the use of contextual information. In: The Second Medway Engineering Conference on Systems: Efficiency, Sustainability and Modelling, Tuesday, 6th June 2017, University of Greenwich, Chatham Maritime ME4 4TB.

Preview	PDF (Author Accepted Manuscript) 17426 MELIS_Reviewing_the_Current_State_of_Machine_Learning_2017.pdf - Accepted Version Download (182kB) | Preview
Preview	PDF (Acceptance Email) 17426 MELIS_Acceptance_Email_2017.pdf - Additional Metadata Download (32kB) | Preview

Abstract

This paper will consider the current state of Machine Learning for Artificial Intelligence, more specifically for applications, such as: Speech Recognition, Game Playing and Image Processing. The artificial world tends to make limited use of context in comparison to what currently happens in human life, while it would benefit from improvements in this area. Additionally, the process of transferring knowledge between application domains is another important area where artificial system can improve. Using context and transferability would have several potential benefits, such as: better ability to function in multiple problem domains, improved understanding of human interaction and stronger grasping of current and potential future situations. While these items are all quite usual to us humans, it is particularly challenging to integrate them into artificial systems, as will be shown within this review. The limitations of our current systems with regards to these topics and the achievable improvements, if these items would be addressed, will also be covered. It is expected that by utilising transferability and/or context, many algorithms in the artificial intelligence field will be able to expand their functionality considerably and should provide for more general purpose learning algorithms.

Item Type:	Conference or Conference Paper (Paper)
Uncontrolled Keywords:	Machine learning, Artificial Intelligence, Transfer learning, Contextual information processing
Subjects:	T Technology > TA Engineering (General). Civil engineering (General)
Faculty / School / Research Centre / Research Group:	Faculty of Engineering & Science Faculty of Engineering & Science > School of Engineering (ENG) Faculty of Engineering & Science > Future Technology and the Internet of Things
Last Modified:	20 Jun 2019 14:04
URI:	http://gala.gre.ac.uk/id/eprint/17426

Actions (login required)

View Item

Downloads