Estimation of ergodicity limits of bag-of-words modeling for guaranteed stochastic convergence. (March 2020)

• Record Type: Journal Article
• Authors: Ghalyan, Ibrahim F. Jasim
• Is Part Of: Pattern recognition. Volume 99(2020:Mar.)
• Publication Date: 2020-03
• Subjects: Bag-of-words -- Ergodicity -- Statistical modeling -- Stochastic process; 00-01 -- 99-00; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2019.107094
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: A dual ergodicity limits-based bag-of-words (DEL-BoW) technique is suggested that guarantees robustness against random initialization, estimates optimal model-order, and achieves enhanced modeling performance. In the DEL-BoW modeling technique, two limits are estimated to the random variable of the performance in order to guarantee ergodicity of the process. The first limit, with a larger radius of convergence, allows to have robustness against random initialization in the modeling process and helps in the estimation of optimal model-order. The second limit, with a smaller radius of convergence, helps in achieving enhanced modeling performance for the optimal model-order. The DEL-BoW modeling technique is applied to Caltech-101, Caltech-256, 15-Scenes, and Flower-102 datasets. Excellent modeling performance is reported when using DEL-BoW modeling technique to the aforementioned datasets. Abstract: This paper suggests an efficient dual ergodicity limits-based bag-of-words (DEL-BoW) modeling technique. The suggested DEL-BoW technique estimates two limits of ergodicity of a discrete random variable (drv) that is formed from the BoW classification performance of multiple runs. The first limit of ergodicity is estimated with a relatively larger ball of convergence to keep the drv shorter. Hence both robustness against random initialization and estimation of the optimal model-order are realized with a reduced number of iterations. Once the optimal model-order is estimated, the radius of ball of convergence is reduced and a second limit of ergodicity is estimated. Reducing the ball of convergence enlarges the size of the considered performance drv that enhances the classification performance. Experiments conducted on Caltech-101, Caltech-256, 15-Scenes, and Flower-102 datasets resulted in classification accuracy of 86.91%, 72.57%, 90.57%, and 90.86%, respectively. Comparison with state-of-the-art techniques shows the excellent performance of the DEL-BoW modeling process.