Multi-perspective creation of diversity for image classification in ensemble learning context

Liu, Han ORCID: https://orcid.org/0000-0002-7731-8258 and Chen, Shyi-Ming 2019. Multi-perspective creation of diversity for image classification in ensemble learning context. Presented at: International Conference on Machine Learning and Cybernetics, Kobe, Japan, 7-10 July 2019. 2019 International Conference on Machine Learning and Cybernetics (ICMLC). 10.1109/ICMLC48188.2019.8949189

Preview

PDF - Accepted Post-Print Version Download (374kB) | Preview

Abstract

Image classification is a special type of classification tasks in the setting of supervised machine learning. In general, in order to achieve good performance of image classification, it is important to select high quality features for training classifiers. However, different instances of images would usually present very diverse features even if the instances belong to the same class. In other words, one types of features may better describe some instances, whereas other instances present more other types of features. The above description would indicate that the adoption of feature selection is likely to result in the case that redundant features are removed from some instances leading to more effective recognition but some important information may get lost from other instances leading to more difficulty in image classification. On the other hand, image features are typically in the form of continuous attributes which can be handled by decision tree learning algorithms in various ways, leading to diverse classifiers being trained. In this paper, we investigate diversified adoption of the C4.5 and KNN algorithms from different perspectives, such as diversified use of features and various ways of handling continuous attributes. In particular, we propose a multi-perspective approach of diversity creation for image classification in the setting of ensemble learning. We compare the proposed approach with those popular algorithms that are used to train classifiers on either a full set of original features or a subset of selected features for image classification. The experimental results show that the performance of image classification is improved through the adoption of our proposed approach of ensemble creation.

Item Type:	Conference or Workshop Item (Paper)
Date Type:	Publication
Status:	Published
Schools:	Computer Science & Informatics
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
ISBN:	9781728128177
Related URLs:	Organisation
Date of First Compliant Deposit:	19 July 2019
Date of Acceptance:	7 May 2019
Last Modified:	07 Dec 2022 13:45
URI:	https://orca.cardiff.ac.uk/id/eprint/124285

Actions (repository staff only)

Edit Item