In this study, empirical prior Dirichlet allocation (epLDA) model that uses latent semantic indexing framework to derive the priors required for topics computation from data is presented. The parameters of the priors so obtained are related to the parameters of the conventional LDA model using exponential function. The model was implemented and tested with benchmarked data and it achieves a prediction accuracy of 92.15%. It was observed that the epLDA model consistently outperforms the conventional LDA model on different datasets with an average percentage accuracy of 6.33%; this clearly demonstrates the advantage of using side information obtained from data for the computation of the mixture components.
Naïve bayes filter is a simple probabilistic filtering method based on Bayes theorem. A crucial problem with the conventional naïve bayes filter is the assumption of uniform priors in the computation of the posterior distribution. For online data such as email environment where the training data are constantly updated so as to outsmart the tricks of spammers, the prior knowledge cannot be uniform. Skewedness in the prior knowledge caused by the updated information has been reported to affect the accuracy and then the effectiveness of the traditional naïve bayes filter. In this study, the skewedness is addressed using complement naïve bayes model. The complement naïve bayes model was implemented and tested on benchmarked data and the result compared with the results obtained with the results obtained from the conventional naïve bayes filter on the same dataset. The complement naïve bayes based filter outperforms the conventional naïve bayes filter by 5.39%.Keywords: Spam, Spam filtering, complement naïve bayes, adaptive filtering, prior, bias, accuracy, filter, adaptive, skewednessVol. 26, No 1, June, 2019
Face images undergo considerable amount of variations in pose, facial expression and illumination condition. This large variation in facial appearances of the same individual makes most Existing Face Recognition Systems (E-FRS) lack strong discrimination ability and timely inefficient for face representation due to holistic feature extraction technique used. In this paper, a novel face recognition framework, which is an extension of the standard (PCA) and (ICA) denoted as two-dimensional Principal Component Analysis (2D-PCA) and two-dimensional Independent Component Analysis (2D-ICA) respectively is proposed. The choice of 2D was advantageous as image covariance matrix can be constructed directly using original image matrices. The face images used in this study were acquired from the publicly available ORL and AR Face database. The features belonging to similar class were grouped and correlation calculated in the same order. Each technique was decomposed into different components by employing multi-dimensional grouped empirical mode decomposition using Gaussian function. The nearest neighbor (NN) classifier is used for classification. The results of evaluation showed that the 2D-PCA method using ORL database produced RA of 92.5%, PCA produced RA of 75.00%, ICA produced RA of 77.5%, 2D-ICA produced RA of 96.00%. However, 2D-PCA methods using AR database produced RA of 73.56%, PCA produced RA of 62.41%, ICA produced RA of 66.20%, 2D-ICA method produced RA of 77.45%. This study revealed that the developed face recognition framework algorithm achieves an improvement of 18.5% and 11.25% for the ORL and AR databases respectively as against PCA and ICA feature extraction techniques. Keywords: computer vision, dimensionality reduction techniques, face recognition, pattern recognition
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.