Δευτέρα, 19 Αυγούστου 2019

A survey of quaternion neural networks

Abstract

Quaternion neural networks have recently received an increasing interest due to noticeable improvements over real-valued neural networks on real world tasks such as image, speech and signal processing. The extension of quaternion numbers to neural architectures reached state-of-the-art performances with a reduction of the number of neural parameters. This survey provides a review of past and recent research on quaternion neural networks and their applications in different domains. The paper details methods, algorithms and applications for each quaternion-valued neural networks proposed.

Incremental methods in face recognition: a survey

Abstract

Face Recognition has rapidly grown as a commercial requirement for a variety of applications in recent years. There are certain situations in which all the face images may not be available before training or the face images may be distributed at geographically apart locations. Incremental face recognition addresses these problems and possesses certain advantages i.e. being time efficient and dynamic model updation allows addition/deletion of samples on the fly. In this paper, a comprehensive review on the Incremental learning algorithms that are aimed at Face Recognition or tested over Face datasets. The contribution of this paper is three-fold: (a) a novel taxonomy of the Incremental methods have been proposed (b) a review of the face datasets used in Incremental face recognition have been carried out and (c) a performance analysis of the Incremental face recognition methods over various face datasets is also presented. Important conclusions have been drawn that will help the researchers in making suitable choices amongst various methods and datasets. This survey shall act as a useful reference to the researchers and practitioners working in incremental face recognition. Furthermore, several viable research directions have been given at the end.

Intuitionistic fuzzy $$\beta $$ β -covering-based rough sets

Abstract

Covering-based rough set is an important extended type of classical rough set model. In this model, concepts are approximated through substitution of a partition in classical rough set theory with a covering in covering-based rough set theory. Various generalized covering-based rough sets have been investigated, however, little work has been done on extending four classical covering-based rough set to intuitionistic fuzzy (IF) settings. In this study, four novel IF covering-based rough set models are developed by combining an IF \(\beta \) -covering with four classical covering-based rough set models. First, we present the concept of IF \(\beta \) -minimal description, and then construct four order relations on IF \(\beta \) approximation space. Second, we propose four IF \(\beta \) -covering-based rough set models and derive that they are generalizations of four existing covering-based rough sets in IF settings. We also discuss the properties of these IF \(\beta \) -covering-based rough sets and reveal their relationships. We use the existing distance between two IF sets to characterize the uncertainty of the presented IF \(\beta \) -covering-based rough sets. Third, we define the reducts of IF \(\beta \) -covering decision systems and examine their discernibility-function-based reduction methods for these IF \(\beta \) -covering-based rough sets. Fourth, we present four optimistic and pessimistic multi-granulation IF \(\beta \) -covering-based rough sets and analyze their properties and uncertainty measures from multi-granulation perspective. Fifth, we study the discernibility-function-based reduction methods for the presented multi-granulation IF \(\beta \) -covering-based rough sets. Finally, we discuss another two neighborhood-based IF covering-based rough sets. This study can provide a covering-based rough set method for acquiring knowledge from IF decision systems.

Feature selection in image analysis: a survey

Abstract

Image analysis is a prolific field of research which has been broadly studied in the last decades, successfully applied to a great number of disciplines. Since the apparition of Big Data, the number of digital images is explosively growing, and a large amount of multimedia data is publicly available. Not only is it necessary to deal with this increasing number of images, but also to know which features extract from them, and feature selection can help in this scenario. The goal of this paper is to survey the most recent feature selection methods developed and/or applied to image analysis, covering the most popular fields such as image classification, image segmentation, etc. Finally, an experimental evaluation on several popular datasets using well-known feature selection methods is presented, bearing in mind that the aim is not to provide the best feature selection method, but to facilitate comparative studies for the research community.

Parameter identification of engineering problems using a differential shuffled complex evolution

Abstract

An accurate mathematical model has a vital role in controlling and synchronization of different systems. But generally in real-world problems, parameters are mixed with mismatches and distortions. In this paper, an improved shuffled complex evolution (SCE) is proposed for parameter identification of engineering problems. The SCE by employing parallel search efficiently finds neighborhoods of the optimal point. So it carries out exploration in a proper way. But its drawback is due to exploitation stages. The SCE cannot converge accurately to an optimal point, in many cases. The current study focuses to overcome this drawback by inserting a shrinkage stage to an original version of SCE and presents a powerful global numerical optimization method, named the differential SCE. The efficacy of the proposed algorithm is first tested on some benchmark problems. After achieving satisfactory performance on the test problems, to demonstrate the applicability of the proposed algorithm, it is applied to ten identification problems includes parameter identification of ordinary differential equations and chaotic systems. Practical experiences show that the proposed algorithm is very effective and robust so that it produces similar and promising results over repeated runs. Also, a comparison against other evolutionary algorithms reported in the literature demonstrates a significantly better performance of our proposed algorithm.

Effective lexicon-based approach for Urdu sentiment analysis

Abstract

The lexicon-based approach is used for sentiment analysis of Urdu. In the lexicon, apart from the traditional approach of having adjectives, nouns and negations we have also included verbs, intensifiers and context-dependent words. An effective Urdu sentiment analyzer is developed that applies rules and make use of this new lexicon and perform Urdu sentiment analysis by classifying sentences as positive, negative or neutral. Evaluating this Urdu sentiment analyzer, by using sentences from Urdu blogs, yields the most promising results so far in Urdu language with 89.03% accuracy with 0.86 precision, 0.90 recall and 0.88 F-measure. Results are evaluated using kappa statistics as well. The comparison with the previous work in Urdu shows that the combination of this Urdu sentiment lexicon and Urdu sentiment analyzer is much more effective than the previous such combinations. The main reason for increased efficiency is the development of wide coverage lexicon and effective handling of negations, intensifiers and context-dependent words by the Urdu sentiment analyzer.

A survey of feature extraction and fusion of deep learning for detection of abnormalities in video endoscopy of gastrointestinal-tract

Abstract

A standard screening procedure involves video endoscopy of the Gastrointestinal tract. It is a less invasive method which is practiced for early diagnosis of gastric diseases. Manual inspection of a large number of gastric frames is an exhaustive, time-consuming task, and requires expertise. Conversely, several computer-aided diagnosis systems have been proposed by researchers to cope with the dilemma of manual inspection of the massive volume of frames. This article gives an overview of different available alternatives for automated inspection, detection, and classification of various GI abnormalities. Also, this work elaborates techniques associated with content-based image retrieval and automated systems for summarizing endoscopic procedures. In this survey, we perform a comprehensive review of feature extraction techniques and deep learning methods which were specifically developed for automatic analysis of endoscopic videos. In addition, we categorize features extraction techniques according to image processing domains and further we classify them based on their visual descriptions. We also review hybrid feature extraction techniques which are developed by the fusion of different kind of basic descriptors. Moreover, this survey covers various endoscopy data-sets available for the bench-marking of vision based algorithms. On the basis of literature, we explain emerging trends in computerized analysis of endoscopy. We also survey important issues, challenges, and future research directions to the development of computer-assisted systems for detection of maladies and interactive surgery in the GI tract.

Recommendation system based on deep learning methods: a systematic review and new directions

Abstract

These days, many recommender systems (RS) are utilized for solving information overload problem in areas such as e-commerce, entertainment, and social media. Although classical methods of RS have achieved remarkable successes in providing item recommendations, they still suffer from many issues such as cold start and data sparsity. With the recent achievements of deep learning in various applications such as Natural Language Processing (NLP) and image processing, more efforts have been made by the researchers to exploit deep learning methods for improving the performance of RS. However, despite the several research works on deep learning based RS, very few secondary studies were conducted in the field. Therefore, this study aims to provide a systematic literature review (SLR) of deep learning based RSs that can guide researchers and practitioners to better understand the new trends and challenges in the field. This paper is the first SLR specifically on the deep learning based RS to summarize and analyze the existing studies based on the best quality research publications. The paper particularly adopts an SLR approach based on the standard guidelines of the SLR designed by Kitchemen-ham which uses selection method and provides detail analysis of the research publications. Several publications were gathered and after inclusion/exclusion criteria and the quality assessment, the selected papers were finally used for the review. The results of the review indicated that autoencoder (AE) models are the most widely exploited deep learning architectures for RS followed by the Convolutional Neural Networks (CNNs) and the Recurrent Neural Networks (RNNs) models. Also, the results showed that Movie Lenses is the most popularly used datasets for the deep learning-based RS evaluation followed by the Amazon review datasets. Based on the results, the movie and e-commerce have been indicated as the most common domains for RS and that precision and Root Mean Squared Error are the most commonly used metrics for evaluating the performance of the deep leaning based RSs.

The dissimilarity approach: a review

Abstract

Dissimilarity representation is a very interesting alternative for the traditional feature space representation when addressing large multi-class problems or even problems with a small number of training samples. This paper describes the existing possibilities in terms of dissimilarity representation through some comprehensive examples. The justification for using such a problem representation strategy is discussed, followed by a complete review of the state-of-art and a critical analysis in which the original purpose of the dissimilarity representation and its perspectives are discussed. Dissimilarity space derived from automatically learned features and the possibility of transiting from one space to another when performing the tasks of the classification process are good examples of promising research directions in this field.

A new graph-preserving unsupervised feature selection embedding LLE with low-rank constraint and feature-level representation

Abstract

Unsupervised feature selection is a powerful tool to process high-dimensional data, in which a subset of features is selected out for effective data representation. In this paper, we proposes a novel robust unsupervised features selection method based on graph-preserving feature selection embedding LLE. Specifically, we integrate the graph matrix learning and the low-dimensional space learning together to identify the correlation among both features and samples from the intrinsic low-dimensional space of original data. Also, the global and local correlation of features have been taken into consideration through the low-rank constraint and the feature-level representation property to find lower-dimensional representation which preserves not only the global and local correlation of features but also the global and local structure of training samples. Furthermore, we propose a new optimization algorithm to the resulting objective function, which iteratively updates the graph matrix and the intrinsic space in order to collaboratively improve each of them. Experimental analysis on 18 benchmark datasets verified that our proposed method outperformed the state-of-the-art feature selection methods in terms of classification and clustering performance.

Δεν υπάρχουν σχόλια:

Δημοσίευση σχολίου