Classification on Grassmann Manifold via Scheiddegger-Watson Distribution using Bayesian Approach

Muhammad Ali, Michael Antolovich

Research output: Book chapter/Published conference paperConference paper

Abstract

Our focus in this paper is a simple Bayesian classification on generalised Scheiddegger-Watson distribution using standard Maximum Likelihood Estimation (MLE). The main barrier in working with Scheiddegger-Watson or matrix variate distributions via standard MLE is the normalising constant that always appears with them. We apply Taylor expansion for approximating the corresponding matrix-based normalising constant and then implement our proposed approach for classification on Grassmann manifold. We then evaluate the effectiveness of our proposed method on real world data against the state of the art recent techniques and show that the proposed approach outperforms or good comparable with them.
Original languageEnglish
Title of host publicationProceedings of the 2016 3rd International Conference on Soft Computing and Machine Intelligence, ISCMI 2016
Place of PublicationUnited States
PublisherIEEE, Institute of Electrical and Electronics Engineers
Pages28-32
Number of pages5
ISBN (Electronic)9781509036967
DOIs
Publication statusPublished - 02 Oct 2017
Event3rd International Conference on Soft Computing & Machine Intelligence : ISCMI 2016 - Flora Grand Hotel, Dubai, United Arab Emirates
Duration: 23 Nov 201625 Nov 2016
Conference number: 3rd
http://www.iscmi.us/ISCMI2016.html

Conference

Conference3rd International Conference on Soft Computing & Machine Intelligence
CountryUnited Arab Emirates
CityDubai
Period23/11/1625/11/16
Internet address

    Fingerprint

Cite this

Ali, M., & Antolovich, M. (2017). Classification on Grassmann Manifold via Scheiddegger-Watson Distribution using Bayesian Approach. In Proceedings of the 2016 3rd International Conference on Soft Computing and Machine Intelligence, ISCMI 2016 (pp. 28-32). [8057433] IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/ISCMI.2016.37