Artificial Intelligence Review

Published by Springer Nature
Online ISSN: 1573-7462
Print ISSN: 0269-2821
Learn more about this page
Recent publications
  • Dingding ChenDingding Chen
  • Ziyu ChenZiyu Chen
  • Yanchen DengYanchen Deng
  • [...]
  • Lulu WangLulu Wang
Asymmetric distributed constraint optimization problems (ADCOPs) are an important framework for multiagent coordination and optimization, where each agent has its personal preferences. However, the existing inference-based complete algorithms that use local eliminations cannot be applied to ADCOPs, as the (pseudo) parents are required to transfer their private functions to their (pseudo) children to perform the local eliminations optimally. Rather than disclosing private functions explicitly to facilitate local eliminations, we solve the problem by enforcing delayed eliminations and propose the first inference-based complete algorithm for ADCOPs, named AsymDPOP. To solve the severe scalability problems incurred by delayed eliminations, we propose to reduce the memory consumption by propagating a set of smaller utility tables instead of a joint utility table, and the computation efforts by sequential eliminations instead of joint eliminations. To ensure the proposed algorithms can scale up to large-scale problems under the limited memory, we combine them with the memory-bounded inference by iteratively propagating the memory-bounded utility tables with the instantiation of cycle-cut (CC) nodes, where we propose to reduce the redundancy in bounded utility iterative propagation by enumerating CC nodes in different branches independently and propagating the utility tables within the memory limit only once. The empirical evaluation indicates that the proposed methods significantly outperform the state-of-the-art as well as the vanilla DPOP with PEAV formulation.
In today’s complex and ever-changing world, Supply Chain Management (SCM) is increasingly becoming a cornerstone to any company to reckon with in this global era for all industries. The rapidly growing interest in the application of Deep Learning (a class of machine learning algorithms) in SCM, has urged the need for an up-to-date systematic review on the research development. The main purpose of this study is to provide a comprehensive vision by reviewing a set of 43 papers about applications of Deep Learning (DL) methods to the SCM, as well as the trends, perspectives, and potential research gaps. This review uses content analysis to answer three research questions namely: 1- What SCM problems have been solved by the use of DL techniques? 2- What DL algorithms have been used to solve these problems? 3- What alternative algorithms have been used to tackle the same problems? And do DL outperform these methods and through which evaluation metrics? This review also responds to this call by developing a conceptual framework in a value-adding perspective that provides a full picture of areas on where and how DL can be applied within the SCM context. This makes it easier to identify potential applications to corporations, in addition to potential future research areas to science. It might also provide businesses a competitive advantage over their competitors by allowing them to add value to their data by analyzing it quickly and precisely.
Knowledge extraction is meant by acquiring relevant information from the unstructured document in natural language and representing them in a structured form. Enormous information in various domains, including agriculture, is available in the natural language from several resources. The knowledge needs to be represented in a structured format to understand and process by a machine for automating various applications. This paper reviews different computational approaches like rule-based and learning-based methods and explores the various techniques, features, tools, datasets, and evaluation metrics adopted for knowledge extraction from the most relevant literature.
The Jaya Algorithm is a relatively new population-based optimization, which has become a progressively valuable tool in swarm intelligence. The Jaya algorithm incorporates the survival of the fittest principle alike evolutionary algorithm by its victorious nature as well as the ideal of an inducement towards a global optimal, which represents its swarm intelligence nature. Nevertheless, it has been applied in various areas of optimization, mainly in engineering practice, which is discussed and abridged based on each problem’s domain.The Jaya optimization’s vast applicability can be explained by its ability to work without any algorithm-specific parameters. The successfully solved problems may also use some of this meta-heuristic’s variants, in which the algorithm has been modified or hybridized. This paper focuses on a comprehensive review, as well as a bibliometric study of the Jaya algorithm, to imply its versatility. Hence, this study is likely to emphasize this optimization’s abilities, inspiring new researchers to make use of this simple and efficient algorithm for problem-solving.
Medical instrument detection is essential for computer-assisted interventions, since it facilitates clinicians to find instruments efficiently with a better interpretation, thereby improving clinical outcomes. This article reviews image-based medical instrument detection methods for ultrasound-guided (US-guided) operations. Literature is selected based on an exhaustive search in different sources, including Google Scholar, PubMed, and Scopus. We first discuss the key clinical applications of medical instrument detection in the US, including delivering regional anesthesia, biopsy taking, prostate brachytherapy, and catheterization. Then, we present a comprehensive review of instrument detection methodologies, including non-machine-learning and machine-learning methods. The conventional non-machine-learning methods were extensively studied before the era of machine learning methods. The principal issues and potential research directions for future studies are summarized for the computer-assisted intervention community. In conclusion, although promising results have been obtained by the current (non-) machine learning methods for different clinical applications, thorough clinical validations are still required.
Over the past decade, unmanned aerial vehicles (UAVs) have demonstrated increasing promise. In this context, we provide a review on swarm intelligence algorithms that play an extremely important role in multiple UAV collaborations. The study focuses on four aspects we consider relevant for the topic: collision avoidance, task assignment, path planning, and formation reconfiguration. A comprehensive investigation of selected typical algorithms that analyses their merits and demerits in the context of multi-UAV collaboration is presented. This research summarises the basic structure of swarm intelligence algorithms, which consists of several fundamental phases; and provides a comprehensive survey of swarm intelligence algorithms for the four aspects of multi-UAV collaboration. Besides, by analysing these key technologies and related applications, the research trends and challenges are highlighted. This broad review is an outline for scholars and professionals in the field of UAV swarms.
The satisfactory performance of the intelligent agents conceived to solve several real-life problems requires efficient decision-making algorithms. To address issues with high state-spaces within reasonable runtime limits, these algorithms are distributed according to some approaches that can be either synchronous or asynchronous, where the former guarantee the same results as their corresponding serial versions through synchronization points, which causes the undesirable effects of communication overhead and idle processors. To mitigate this, the asynchronous approaches reduce the message exchanges in such a way as to accelerate the runtime without too much compromising the response accuracy. The challenge of enhancing the minimax technique through pruning makes Alpha–Beta a relevant case study in parallelism research. Young Brothers Wait Concept (YBWC) and Asynchronous Parallel Hierarchical Iterative Deepening (APHID) are highlighted among the existing Alpha–Beta distributions. Knowing that APHID proved to be more suitable than YBWC to operate in distributed memory and that shared memory architectures are scarcely available due to their high costs, the primary motivation here is to implement the Asynchronous Distributed Alpha–Beta Algorithm (ADABA), which increases the accuracy and performance of APHID through the enhancement of the slaves’ task ordering policies, the communication process between the processors and the window’s updating strategy. Experiments fulfilled through tournaments involving ADABA-based and APHID-based Checkers agents proved that the player based on the best ADABA version reached, approximately, a victory rate 95% superior and a runtime two times faster than the APHID-based player, keeping the same response accuracy level of its opponent.
Deep learning is a popular tool for image recognition due to its good feature learning capability. However, the training of most deep networks suffers from high computing costs and is time-consuming in image recognition. In this paper, we propose a broad-based ensemble learning model (BELM) that aims to provide a fast and efficient recognition approach for moderate-large scale image sets on an ordinary computer. This model is constructed in the form of a flat network, of which the flatted input consists of different feature nodes mapped from original inputs. The structure is expanded in a wide fashion in feature nodes and broad incremental learning algorithms are developed for the dynamic updating of feature nodes when the system deems to be expanded. Also, Lasso sparse autoencoder is considered for feature nodes to achieve compact and sparse features. Compared with most of the existing state-of-the-art networks, the structure of the proposed BELM is uncomplicated and few parameters need to be adjusted. Extensive experimental results on the classical data sets of the handwritten digital database (MNIST) and the object recognition dataset (NORB) demonstrate the effectiveness of the proposed model.
Plants can be seen everywhere in daily life and are closely connected with our lives. The recognition and classification of plants are of great significance to ecological and environmental protection. Traditional plant identification methods are complex, and experts cannot classify multiple plant species quickly. More and more researchers pay attention to image processing and pattern recognition and use them to identify and classify plant leaves quickly. Based on this, this paper summarizes and classifies the methods of plant leaf recognition in recent years. First, we analyze these studies and classify them using different features and classifiers, such as shape, texture, color features, support vector machines, K nearest neighbors, convolutional neural networks, and so on. Secondly, compare the recognition results of plant leaf recognition methods under different datasets. Finally, the recognition of plant leaves is summarized, and future research and development have prospected.
A hybrid method for energy management on grid-connected MG system is proposed under this manuscript. Grid-connected MG system takes photovoltaic (PV), wind turbine (WT), battery. The proposed system is an integration of seagull optimization algorithm (SOA) and the radial basic functional neural network (RBFNN), thus it is named SOA-RBFNN. Here, in the grid-connected microgrid configuration, the necessary load demand is always monitored with RBFNN methodology. SOA optimizes the perfect match of the MG taking into account the predictable load requirement. The fuel cost, together with the power variation per hour of the electric grid, the operation and maintenance cost of microgrid system linked with grid, is described. The proposed model runs on the MATLAB/Simulink workstation and efficiency is investigated using existing techniques as AGO-RNN and MBFA-ANN. Statistical analysis, elapsed time, modeling metrics, and determination of optimal sample size for adjustment and validation of proposed and existing technique are evaluated. The efficiency values on the 100, 200, 500, and 1000 trails are 99.7673%, 99.7609%, 99.9099%, and 99.9373%.
In the present situation, environment is rapidly polluted by the manufacturing of non-eco-friendly products and carbon emission from production industries. So, to control this pollution as well as carbon emission, everyone should be aware to use eco-friendly products and reduce carbon emission. Motivating on this topic, a model on green manufacturing system has been formulated where the produced items are eco-friendly and have a fixed lifetime. Also, in this model, the green level of the produced items enhances the demand of the customers. The objective of this work is to determine the green level of the product and business period by maximizing the average profit of the system. To solve the corresponding maximization problems, a hybrid tournament differential evolution (TDE) algorithm is applied to obtain the best-found solutions along with the average profit of the system. To check the validity of the proposed model, a numerical example is considered and solved by the different variants of the said algorithm. Also, the simulated results obtained from the different variants of TDE algorithm are compared with the simulated results obtained from some of the existing algorithms reported in the literature. Then to test the performance and efficiency of the said hybrid algorithm, statistical comparisons, statistical tests are performed. Finally, sensitivity analyses are carried out in order to examine the effects of changes of parameters involved in the model on the average profit, green level of the product, business and production periods.
The Big Video Data generated in today's smart cities has raised concerns from its purposeful usage perspective, where surveillance cameras, among many others are the most prominent resources to contribute to the huge volumes of data, making its automated analysis a difficult task in terms of computation and preciseness. Violence Detection (VD), broadly plunging under Action and Activity recognition domain, is used to analyze Big Video data for anomalous actions incurred due to humans. The VD literature is traditionally based on manually engineered features, though advancements to deep learning based standalone models are developed for real-time VD analysis. This paper focuses on overview of deep sequence learning approaches along with localization strategies of the detected violence. This overview also dives into the initial image processing and machine learning-based VD literature and their possible advantages such as efficiency against the current complex models. Furthermore,the datasets are discussed, to provide an analysis of the current models, explaining their pros and cons with future directions in VD domain derived from an in-depth analysis of the previous methods.
Feature selection facilitates intelligent information processing, and the unsupervised learning of feature selection has become important. In terms of unsupervised feature selection, the Laplacian score (LS) provides a powerful measurement and optimization method, and good performance has been achieved using the recent forward iterative Laplacian score (FILS) algorithm. However, there is still room for advancement. The aim of this paper is to improve the FILS algorithm, and thus, feature significance (SIG) is mainly introduced to develop a high-quality selection method, i.e., the incremental forward iterative Laplacian score (IFILS) algorithm. Based on the modified LS, the metric difference in the incremental feature process motivates SIG. Therefore, SIG offers a dynamic characterization by considering initial and terminal states, and it promotes the current FILS measurement on only the terminal state. Then, both the modified LS and integrated SIG acquire granulation nonmonotonicity and uncertainty, especially on incremental feature chains, and the corresponding verification is achieved by completing examples and experiments. Furthermore, a SIG-based incremental criterion of minimum selection is designed to choose optimization features, and thus, the IFILS algorithm is naturally formulated to implement unsupervised feature selection. Finally, an in-depth comparison of the IFILS algorithm with the FILS algorithm is achieved using data experiments on multiple datasets, including a nominal dataset of COVID-19 surveillance. As validated by the experimental results, the IFILS algorithm outperforms the FILS algorithm and achieves better classification performance.
Comparison of the granularity reduction efficiency on different size of granularity
Comparison of the granularity reduction efficiency on different size of dataset
In order to overcome the limitation of low efficiency of existing granularity reduction algorithms in multi-granulation rough sets, based on matrix method, a fast granularity reduction algorithm is proposed and the time complexity is \(O({|U |}^{2} \cdot |A |+ |U|\cdot {|A |}^{2})\). First, the definitions of positive region matrix and granularity column matrix of multi-granulation space are proposed. Second, through the quantity product of these two matrices, the definition of positive region column matrix is presented. Based on the positive region column matrix, cut matrix and matrix norm are defined, respectively. Third, the matrix-based calculation methods of multi-granulation approximation quality and granularity significance are proposed. Finally, a heuristic rule is designed according to the granularity significance, and a matrix-based fast granularity reduction algorithm is proposed. Experimental results demonstrate the effectiveness of the proposed methods.
Process of the proposed method
The accumulated historical data are beneficial for generating solutions that are more satisfactory to decision makers because their preferences and experience are characterized by historical data. However, this might be infeasible when only few data are available. Suppose that the few data are collected from a domain called the target domain. There may be some domains correlated to the target domain, which are called source domains. The data from source domains might be useful for helping generate solutions to the problem in the target domain. Following this idea, this paper proposes a cross-domain decision making method based on the combination of TrAdaBoost, an instance-based transfer learning method, and a decision making method in the context of the evidential reasoning approach. This may be the first attempt to combine transfer learning with a decision making method to help generate high-quality solutions satisfactory to decision makers when only few data are available for the problem in the target domain. A data selection strategy is designed to increase the similarity between the data from source and target domains and a weight initialization strategy is designed based on the available gold standards. The two strategies are intended for improving the performance of the proposed method. With the two strategies, the process of the proposed method is presented. The effectiveness of the proposed method is validated by its application in helping diagnose breast lesions with the diagnostic data of five radiologists collected from a tertiary hospital located in Hefei, Anhui, China.
Automatically understanding the content of medical images and delivering accurate descriptions is an emerging field of artificial intelligence that combines skills in both computer vision and natural language processing fields. Medical image captioning is involved in various applications related to diagnosis, treatment, report generation and computer-aided diagnosis to facilitate the decision making and clinical workflows. Unlike generic image captioning, medical image captioning highlights the relationships between image objects and clinical findings, which makes it a very challenging task. Although few review papers have already been published in this field, their coverage is still quite limited and only particular problems are addressed. This motivates the current paper where a rapid review protocol was adopted to review the latest achievements in automatic medical image captioning from the medical domain perspective. We aim through this review to provide the reader with an up-to-date literature in this field by summarizing the key findings and approaches in this field, including the related datasets, applications and limitations as well as highlighting the main competitions, challenges and future directions.
The overall framework for population reduction with individual similarity
The pseudocode for DE/rand/1/bin with the PRS
The dynamic change of population obtained by PRS-DE/rand/1 and DE/rand/1 on the Rosenbrock function with 2D\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$2D$$\end{document}. a Surface and contour plot, b Contour for distribution of NPG=40\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$NP_{G} = 40$$\end{document} individuals at generation G=0\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G = 0$$\end{document}, cNPG=33\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$NP_{G} = 33$$\end{document} of PRS-DE/rand/1 at G=8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G = 8$$\end{document}, dNPG=20\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$NP_{G} = 20$$\end{document} of PRS-DE/rand/1 at G=15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$G = 15$$\end{document}
Choosing the appropriate population size for differential evolution (DE) is still a challenging task. Too large population size leads to slow convergence, while too small population size causes premature convergence and stagnation. To solve this problem, a population reduction with individual similarity (PRS) for DE is proposed in this paper. In the PRS, a linear differential decrease method is used to automatically determine the population size required in each generation. At the same time, the current population is divided into two subgroups with equal sizes according to individual similarity, and the individuals that need to be removed are determined from the subgroup with the lowest individual similarity in an effective manner, and thus the convergence is further accelerated without affecting the population diversity. In addition, an elite-oriented strategy is utilized to replace the random selection of individuals in the original mutation strategy of DE, which provides constructive guidance for individual evolution and improves the convergence quality. Five basic DE and six advanced DE algorithms are used to evaluate the effect of PRS, and it is further compared with four improved DE algorithms with population reduction strategy. The experimental results on CEC 2014 benchmark functions show that the proposed PRS can effectively enhance the performance of these five basic DE and six advanced DE algorithms, and is better than the four population reduction strategies.
Federated learning (FL) is a promising framework for distributed machine learning that trains models without sharing local data while protecting privacy. FL exploits the concept of collaborative learning and builds privacy-preserving models. Nevertheless, the integral features of FL are fraught with problems, such as the disclosure of private information, the unreliability of uploading model parameters to the server, the communication cost, etc. Blockchain, as a decentralized technology, is able to improve the performance of FL without requiring a centralized server and also solves the above problems. In this paper, a systematic literature review on the integration of Blockchain in federated learning was considered with the analysis of the existing FL problems that can be compensated. Through carefully screening, most relevant studies are included and research questions cover the potential security and privacy attacks in traditional federated learning that can be solved by blockchain as well as the characteristics of Blockchain-based FL. In addition, the latest Blockchain-based approaches to federated learning have been studied in-depth in terms of security and privacy, records and rewards, and verification and accountability. Furthermore, open issues related to the combination of Blockchain and FL are discussed. Finally, future research directions for the robust development of Blockchain-based FL systems are proposed.
Portfolio optimization has always been a challenging proposition in finance and management. Portfolio optimization facilitates in selection of portfolios in a volatile market situation. In this paper, different classical, statistical and intelligent approaches employed for portfolio optimization and management are reviewed. A brief study is performed to understand why portfolio is important for any organization and how recent advances in machine learning and artificial intelligence can help portfolio managers to take right decisions regarding allotment of portfolios. A comparative study of different techniques, first of its kind, is presented in this paper. An effort is also made to compile classical, intelligent, and quantum-inspired techniques that can be employed in portfolio optimization.
Hierarchical relations between entity and aspects
Overview of the presented methods for aspect detection. Due to the lack of comparative results, we do not present the topic modeling techniques in the following sections
Sentiment analysis is an important tool to automatically understand the user-generated content on the Web. The most fine-grained sentiment analysis is concerned with the extraction and sentiment classification of aspects and has been extensively studied in recent years. In this work, we provide an overview of the first step in aspect-based sentiment analysis that assumes the extraction of opinion targets or aspects. We define a taxonomy for the extraction of aspects and present the most relevant works accordingly, with a focus on the most recent state-of-the-art methods. The three main classes we use to classify the methods designed for the detection of aspects are pattern-based, machine learning, and deep learning methods. Despite their differences, only a small number of works belong to a unique class of methods. All the introduced methods are ranked in terms of effectiveness. In the end, we highlight the main ideas that have led the research on this topic. Regarding future work, we deemed that the most promising research directions are the domain flexibility and the end-to-end approaches.
Fuzzy clustering has been useful in capturing the uncertainty present in the data during clustering. Most of the c-Means algorithms such as FCM (Fuzzy c-Means), IFCM (Intuitionistic Fuzzy c-Means), and the recently reported PIFCM (Probabilistic Intuitionistic Fuzzy c-means) randomly initialize cluster centroids. Performance of these techniques is very reliant on the initialized cluster centroids. So, a good initialization technique can significantly affect the cluster formation. Recently, density-based initialization technique for FCM (DFCM) was proposed, which initializes datapoints with high density as cluster centroids. In DFCM, points within some distance contribute in the density of the data points. In this paper, we propose a new way to compute fuzzy density of datapoints based on the distance measure. Uncertainty can be better captured by intuitionistic fuzzy set (IFS) and interval-valued IFS (IVIFS). Thus, we propose a new density-based initialization technique for IFCM, called ‘Density based Intuitionistic Fuzzy c-Means (DIFCM) Algorithm’. The proposed DIFCM has been further developed for IVIFS, which we term ‘Interval-valued Density based Intuitionistic Fuzzy c-Means (IVDIFCM) Algorithm’, is also introduced in this paper. PIFCM incorporates probabilistic weights between membership, non-membership and hesitancy component. In this paper, we also introduce the density based initialized cluster centroids for PIFCM algorithm to propose the ‘Density Based Probabilistic Intuitionistic Fuzzy c-Means (DPIFCM) Algorithm’. There were many clustering approaches based on IFSs but there do not exist any literature review on the IFS based clustering approaches. Therefore, this article also provides a detailed review of the recently proposed clustering algorithms based on IFS theory. Experiments over various UCI datasets proves that our proposed algorithms has better clustering results than their existing counterparts.
The unprecedented rise in research interest in artificial intelligence (AI) and related areas, such as computer vision, machine learning, robotics, and cognitive science, during the last decade has fuelled the development of software platforms that can simulate embodied agents in 3D virtual environments. A simulator that closely mimics the physics of a real-world environment with embodied agents can allow open-ended experimentation, and can circumvent the need for real-world data collection, which is time-consuming, expensive, and in some cases, impossible without privacy invasion, thereby playing a significant role in progressing AI research. In this article, we review 22 simulation platforms reported in the literature. We classify them based on visual environment and physics. We present a comparison of these simulators based on their properties and functionalities from a user’s perspective. While no simulator is better than the others in all respects, a few stand out based on a rubric that encompasses the simulators’ properties, functionalities, availability and support. This review will guide users to choose the appropriate simulator for their application and provide a baseline to researchers for developing state-of-the-art simulators.
The past decade has witnessed the adoption of artificial intelligence (AI) in various applications. It is of no exception in the area of prognostics and health management (PHM) where the capacity of AI has been highlighted through numerous studies. In this paper, we present a comprehensive review of AI-based solutions in engineering PHM. This review serves as a guideline for researchers and practitioners with varying levels of experience seeking to broaden their know-how about AI-based PHM. Specifically, we provide both a broad quantitative analysis and a comprehensive qualitative examination of the roles of AI in PHM. The quantitative analysis offers an insight into the research community’s interest in AI-based approaches, focusing on the evolution of research trends and their developments in different PHM application areas. The qualitative survey gives a complete picture on the employment of AI in each stage of the PHM process, from data preparation to decision support. Based on the strengths and weaknesses of existing methods, we derive a general guideline for choosing proper techniques for each specific PHM task, aiming to level up maintenance practitioners’ efficiency in implementing PHM solutions. Finally, the review discusses challenges and future research directions in the development of autonomous intelligent PHM solutions.
CS approaches: a conventional CS; b DL for CS reconstruction; c DL for CS sampling and reconstruction; d DL for CS direct inference
Number of scientific papers on the topic CS–DL published between 2015 and 2020 (Data from Google Scholar using the search terms “deep learning” and “compressed sensing”)
A schematic illustration of the survey’s organization
CS signal reconstruction using an iterative algorithm unrolled over a deep neural network. The CS iterative algorithm’s iteration step h is executed a number of times, resulting in the network layers h1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$h_1$$\end{document}, h2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$h_2$$\end{document}, ⋯\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\cdots$$\end{document}, hn\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$h_n$$\end{document}
CS signal reconstruction where a traditional iterative algorithm is fully (or partly) replaced using a deep neural network. The CS iterative algorithm’s iteration step h, executed a number of times, is implemented using a certain neural network architecture
Compressive sensing (CS) is a mathematically elegant tool for reducing the sensor sampling rate, potentially bringing context-awareness to a wider range of devices. Nevertheless, practical issues with the sampling and reconstruction algorithms prevent further proliferation of CS in real world domains, especially among heterogeneous ubiquitous devices. Deep learning (DL) naturally complements CS for adapting the sampling matrix, reconstructing the signal, and learning from the compressed samples. While the CS–DL integration has received substantial research interest recently, it has not yet been thoroughly surveyed, nor has any light been shed on practical issues towards bringing the CS–DL to real world implementations in the ubiquitous computing domain. In this paper we identify main possible ways in which CS and DL can interplay, extract key ideas for making CS–DL efficient, outline major trends in the CS–DL research space, and derive guidelines for the future evolution of CS–DL within the ubiquitous computing domain.
Havoc, brutality, economic breakdown, and vulnerability are the terms that can be rightly associated with COVID-19, for the kind of impact it is having on the whole world for the last two years. COVID-19 came as a nightmare and it is still not over yet, changing its form factor with each mutation. Moreover, each unpredictable mutation causes more severeness. In the present article, we outline a decision support algorithm using Generalized Trapezoidal Intuitionistic Fuzzy Numbers (GTrIFNs) to deal with various facets of COVID-19 problems. Intuitionistic fuzzy sets (IFSs) and their continuous counterparts, viz., the intuitionistic fuzzy numbers (IFNs), have the flexibility and effectiveness to handle the uncertainty and fuzziness associated with real-world problems. Although a meticulous amount of research works can be found in the literature, a wide majority of them are based mainly on normalized IFNs rather than the more generalized approach, and most of them had several limitations. Therefore, we have made a sincere attempt to devise a novel Similarity Measure (SM) which considers the evaluation of two prominent features of GTrIFNs, which are their expected values and variances. Then, to establish the superiority of our approach we present a comparative analysis of our method with several other established similarity methods considering ten different profiles of GTrIFNs. The proposed SM is then validated for feasibility and applicability, by elaborating a Fuzzy Multicriteria Group Decision Making (FMCGDM) algorithm and it is supportedby a suitable illustrative example. Finally, the proposed SM approach is applied to tackle some significant concerns due to COVID-19. For instance, problems like the selection of best medicine for COVID-19 infected patients; proper healthcare waste disposal technique; and topmost government intervention measures to prevent the COVID-19 spread, are some of the burning issues which are handled with our newly proposed SM approach. Graphical abstract
Multiple taxonomies-part 1
Multiple taxonomies-part 2
Since its emergence in the 1960s, Arti�cial Intelligence (AI) has grown to conquer most technology products and their �elds of application. Machine learning, as a major part of the current AI solutions, can learn from the data and through experience to reach high performance on various tasks. This growing success of AI algorithms has led to a need for interepretability to understand opaque models such as deep learning networks. Various requirements have been raised from di�erent domains, together with numerous tools to debug, justify outcomes, and establish safety, fairness and reliability of the models. This variety of tasks has led to inconsistencies in the terminology with, for instance, terms such as interpretable, explainable and transparent being often used interchangeably in methodology papers. These words, however, convey di�erent meanings and are \weighted" di�erently across domains, for example in the technical and social sciences. In this paper, we propose an overarching terminology of interpretability of AI systems that can be referred to by the technical developers as much as by the social sciences community to pursue clarity and e�ciency in the de�nition of regulations for ethical and reliable AI development. We show how our de�nitions di�er from the ones in previous taxonomies and how they apply with high versatility to several domains and use cases, proposing a - highly needed - standard for the communication among interdisciplinary areas of AI.
The cardiovascular disease is one of the most serious health problems around the world. Traditionally, detection of cardiac arrhythmia based on the visual inspection of electrocardiography (ECG) signals by cardiologists is tedious, laborious and subjective. To overcome such disadvantages, numerous arrhythmia detection techniques including signal processing and machine learning tools have been developed. However, there still remain the problems of automatic detection with high efficiency and accuracy in distinguishing different myocardial dysfunctions through ECG signals. In this study we propose a novel technique for automatic detection of cardiac arrhythmia in one-lead ECG signals based upon complete ensemble empirical mode decomposition (CEEMD), discrete wavelet transform (DWT), phase space reconstruction (PSR) and neural networks. First, ECG signals are decomposed into a series of Intrinsic Mode Functions (IMFs) by using the CEEMD method without the preprocessing of QRS detection. The IMF6 and IMF7 of the ECG signals are extracted, which contain the majority of the ECG signals’ energy and are considered to be the predominant IMFs. Second, four levels DWT is employed to decompose the predominant IMFs into different frequency bands, in which third-order Daubechies (db3) wavelet function is selected as reference variable for analysis. Third, phase space of the reference variable is reconstructed based on db3, in which the properties associated with the nonlinear ECG system dynamics are preserved. Three-dimensional (3D) PSR together with Euclidean distance (ED) has been utilized to derive features, which demonstrate significant difference in ECG system dynamics between normal versus abnormal individual heartbeats. Fourth, neural networks are then used to model, identify and classify ECG system dynamics between normal (healthy) and arrhythmia ECG signals. Finally, experiments are carried out on the MIT-BIH arrhythmia database to assess the effectiveness of the proposed method, in which 436 ECG signal fragments for one lead (MLII) from 28 persons of five classes of heart beats were extracted. By using the 10-fold cross-validation style, the achieved average classification accuracy is reported to be \(98.81\%\). Compared with various state-of-the-art methods, the results demonstrate superior performance and the proposed method can serve as a potential candidate for the automatic detection of myocardial dysfunction in the clinical application.
Trends of the publications containing “interpretable AI" or “explainable AI" as keywords
Graphical representation of Artificial Intelligence, Machine Learning, and Deep Learning adapted from
Differences of definitions in other domains than ML development. In this diagram, interpretable is equated to explainable since most of the social domains equate the two terms for simplicity
Multiple taxonomies-part 1
Multiple taxonomies-part 2
Since its emergence in the 1960s, Artificial Intelligence (AI) has grown to conquer many technology products and their fields of application. Machine learning, as a major part of the current AI solutions, can learn from the data and through experience to reach high performance on various tasks. This growing success of AI algorithms has led to a need for interpretability to understand opaque models such as deep neural networks. Various requirements have been raised from different domains, together with numerous tools to debug, justify outcomes, and establish the safety, fairness and reliability of the models. This variety of tasks has led to inconsistencies in the terminology with, for instance, terms such as interpretable , explainable and transparent being often used interchangeably in methodology papers. These words, however, convey different meanings and are “weighted" differently across domains, for example in the technical and social sciences. In this paper, we propose an overarching terminology of interpretability of AI systems that can be referred to by the technical developers as much as by the social sciences community to pursue clarity and efficiency in the definition of regulations for ethical and reliable AI development. We show how our taxonomy and definition of interpretable AI differ from the ones in previous research and how they apply with high versatility to several domains and use cases, proposing a—highly needed—standard for the communication among interdisciplinary areas of AI.
Concepts based on psychology fit well with current research trends related to robotics and artificial intelligence. Biology-inspired cognitive architectures are extremely useful in building agents and robots, and this is one of the most important challenges of modern science. Therefore, the widely viewed and far-reaching goal of systems research and engineering is virtual agents and autonomous robots that mimic human behavior in solving known and unknown problems. The article proposes, at a high level of generality, an operational cybernetic model of the human mind, developed with the use of carefully selected ideas taken from psychological knowledge. In particular, the work combines extensive knowledge drawn from both the theory of developmental cognitive psychology and the theory of motivation. The proposed mathematically developed operating blocks create a coherent and functional decision-making system containing all the elements necessary in autonomous robotics. The ISD system is under development. There is still a long way to go to full validation. However, as shown in several articles, the basic subsystems of the ISD system, i.e. motivational and emotional, have already been positively verified in operation. The overall purpose of this article is to show a blueprint of the overall concept of the entire ISD.
In recent years, a large number of manipulator robots have been deployed to replace or assist humans in many repetitive and dangerous tasks. Yet, these robots have complex mechanisms, resulting in their non-linearity of kinematics and dynamics as well as intensive computations. Therefore, relying on soft computing techniques are a common and alternative key to model and control these systems. In particular, fuzzy logic approaches have proven to be simple, efficient, and superior to relevant well-known methods and have sparked greater interest in robotic applications. To help researchers meet their needs easily and quickly in finding relevant research works on fuzzy-based solutions, this article adapted to provide an in-depth review of the currently updated fuzzy logic approaches for collision-free path planning of serial manipulator robots operating in complex and cluttered workspaces. In addition to a comprehensive description of fuzzy hybridization with other artificial intelligence techniques description. Further, this article attempts to present the main solutions with a summary and visualization of all basic approaches that path-planning problems may subtend in the decision-making process. Finally, the paper suggests some potential challenges and explores research issues for future work.
The task of anomaly detection has recently gained much attention in the field of visual surveillance. Video surveillance data is often available in large quantities, but manual annotation of activities in video segments is tedious. Anomaly detection plays a crucial role in various indoor and outdoor surveillance applications. Video anomaly detection is highly challenging and provides a lot of scope and demand for improving detection performance in real-time scenarios. Recently, deep learning-based approaches are promising to detect single-scene video anomalies in real-time. This work starts by highlighting the over-view of deep learning-based video anomaly detection. A thematic taxonomy that includes four major categories and several sub-categories is presented. State-of-the-art deep learning approaches under these categories are reviewed. In addition, few recent one class model based deep learning approaches are evaluated and analyzed in terms of performance. Out of the approaches presented, Generative Adversarial Network (GAN) and Adversarial Autoencoder-based approaches provide a better detection rate. A few important directions are outlined for further research in the field of video surveillance applications.
Retinal fundus image analysis (RFIA) is frequently used in diabetic retinopathy (DR) scans to determine the risk of blindness in diabetic patients. Ophthalmologists receive support from various RFIA programs to cope with the detection of visual impairments. In this article, active deep learning (ADL) using new multi-layer architecture for automatic recognition of DR stages is presented. In order to facilitate the detection of retinal lesions in the ADL system preprocessing, the image is segmented using the artificial bee colony (ABC) algorithm with a threshold value determined according to the results of the image histogram. Besides, a tag-efficient convolutional neural networks (CNN) architecture known as ADL-CNN has been developed to automatically extract segmented retinal features. This model has a two-stage process. In the first, images are selected to learn simple or complex retinal features using basic accuracy labels in the training examples. Second, useful masks are provided with key lesion features and segment areas of interest within the retinal image. Performance evaluation of the proposed ADL-CNN model is made by comparing the most advanced methods using the same dataset. The efficiency of the system is made by measuring statistical metrics such as classification accuracy (ACC), sensitivity (SE), specificity (SP), and F-measure. The ADL-CNN model applied to the EyePacs dataset containing 35,122 retinal images yielded 99.66% ACC, 93.76% SE, 96.71% SP, and 94.58% F-measure. In this respect, it can be said that the proposed method shows high performance in detecting DR lesions from various fundus images and determining the severity level.
Scheduling has an immense effect on various areas of human lives, be it though its application in manufacturing and production industry, transportation, workforce allocation, or others. The unrelated parallel machines scheduling problem (UPMSP), which is one of the various problem types that exist, found its application in many areas like manufacturing and distributed computing. Due to the complexity of the problem, heuristic and metaheuristic methods have dominantly been applied for solving it. Although this problem variant did not receive much attention as other models, recent years saw the increase of research dealing with the UPMSP. During that time, different problem variants, solution methods, and interesting research directions were considered. However, no study provided a systematic overview of the research in which heuristic methods are applied for solving the UPMSP. This comes as a problem since it is becoming difficult to keep track of all the relevant research directions and solution methods considered for this problem. Therefore, the goal of this study is to provide an extensive literature review on the application of heuristic and metaheuristic methods for solving the UPMSP. Each reviewed study is briefly described based on the considered problem and solution method. Additionally, studies dealing with similar problems are grouped together to outline the evolution of the research, and possible areas where further research can be carried out. All studies were systematised and classified into several categories to allow for an easy overview of different problem and solution variants. Finally, recent research trends and possible future directions are also outlined.
Knowledge discovery and evaluation is a challenging but rewarding process of obtaining available information automatically from database. Due to the heterogeneity of the collected data, the connotative knowledge has the characteristics of uncertainty, random occurrence and variable scale. Therefore, an unsupervised knowledge discovery and variable scale evaluation model is presented in this paper based on a new multi-feature fusion method. Firstly, point at the multiple information features, an amplitude-frequency-shape based state description form is proposed in this paper. It could analyze the time series from the aspects of energy, phase, and knowledge similarity. In view of the variable number and scale of knowledge fragments, a piecewise linear segmentation criterion is put forward based on the complexity and accuracy of information representation. Then a model free knowledge discovery framework without samples labels is constructed to discover the knowledge quickly and effectively. Aimed at the variable knowledge scale, a variable scale evaluation method is first proposed to distinguish the multi-scale decision-making knowledge based on the indicators of system stability and security. It could optimize the knowledge base and guide the decision-making process. The experimental results on heterogeneous activity datasets indicate that the proposed method here could generally analysis the time series state and discover the knowledge efficiently from massive data. In addition, the knowledge discovery and evaluation at a continuous decision system show that the proposed framework could meet the needs of knowledge discovery in complex environment and effectively distinguish the knowledge to provide strong support for establishing a credible decision-making system.
Dialogue systems are a popular natural language processing (NLP) task as it is promising in real-life applications. It is also a complicated task since many NLP tasks deserving study are involved. As a result, a multitude of novel works on this task are carried out, and most of them are deep learning based due to their outstanding performance. In this survey, we mainly focus on the deep learning based dialogue systems. We comprehensively review state-of-the-art research outcomes in dialogue systems and analyze them from two angles: model type and system type. Specifically, from the angle of model type, we discuss the principles, characteristics, and applications of different models that are widely used in dialogue systems. This will help researchers acquaint these models and see how they are applied in state-of-the-art frameworks, which is rather helpful when designing a new dialogue system. From the angle of system type, we discuss task-oriented and open-domain dialogue systems as two streams of research, providing insight into the hot topics related. Furthermore, we comprehensively review the evaluation methods and datasets for dialogue systems to pave the way for future research. Finally, some possible research trends are identified based on the recent research outcomes. To the best of our knowledge, this survey is the most comprehensive and up-to-date one at present for deep learning based dialogue systems, extensively covering the popular techniques. We speculate that this work is a good starting point for academics who are new to the dialogue systems or those who want to quickly grasp up-to-date techniques in this area.
Students typically do not have practical tools to help them choose their target universities to apply. This work proposes a comprehensive analytics framework as a decision support tool that assists students in their admission process. As an essential element of the developed framework, a prediction procedure is developed to precisely determine the student's chance of admission to each university using various machine learning methods. It is concluded that random forest combined with kernel principal component analysis outperforms other prediction models. Besides, an online survey is built to elicit the utility of the student regarding each university. A mathematical programming model is then proposed to determine the best universities to apply among the candidates considering the probable limitations; the most important is the student's budget. The model is also extended to consider multiple objectives for making decisions. Last, a case study is provided to show the practicality of the developed decision support tool.
The Probabilistic Hesitant Fuzzy Sets (PHFS) based on the hesitant fuzzy sets has been paid great attention. Though numerous methods have been applied in this environment since the PHFS has been introduced, there are still new fields to be explored. The EDAS method which is the abbreviation of the evaluation based on distance from average solution is one of the practical methods in circumstances which is with contradictory attributes. Considering the uncertain character of the PHF condition and the psychological factors which influence decision makers’ behaviors such as the character and risk reference, the probabilistic hesitant fuzzy EDAS integrating with cumulative prospect theory (PHF-CPT-EDAS) is built for multiple attributes group decision making (MAGDM) problem. Meanwhile, the information of entropy is also utilized to calculate the unknown weighting vector of attributes. At last, we utilize two case studies to compare the designed method with other MADM methods. Through this article, we learn that the PHF-CPT-EDAS method is effective and stable to solve the MAGDM issues.
Researchers are defining new types of interactions between humans and machine learning algorithms generically called human-in-the-loop machine learning. Depending on who is in control of the learning process, we can identify: active learning, in which the system remains in control; interactive machine learning, in which there is a closer interaction between users and learning systems; and machine teaching, where human domain experts have control over the learning process. Aside from control, humans can also be involved in the learning process in other ways. In curriculum learning human domain experts try to impose some structure on the examples presented to improve the learning; in explainable AI the focus is on the ability of the model to explain to humans why a given solution was chosen. This collaboration between AI models and humans should not be limited only to the learning process; if we go further, we can see other terms that arise such as Usable and Useful AI. In this paper we review the state of the art of the techniques involved in the new forms of relationship between humans and ML algorithms. Our contribution is not merely listing the different approaches, but to provide definitions clarifying confusing, varied and sometimes contradictory terms; to elucidate and determine the boundaries between the different methods; and to correlate all the techniques searching for the connections and influences between them.
Brain is an amazing organ that controls all activities of a human. Any abnormality in the shape of anatomical regions of the brain needs to be detected as early as possible to reduce the mortality rate. It is also beneficial for treatment planning and therapy. The most crucial task is to isolate abnormal areas from normal tissue regions. To identify abnormalities in the earlier stage, various medical imaging modalities were used by medical practitioners as part of the diagnosis. Magnetic Resonance Imaging (MRI) is a non-invasive diagnostic tool used for analyzing the internal structures owing to its capability to provide images with high resolution and better contrast for soft tissues. This survey focuses on studies done in brain MRI. Manual segmentation of abnormal tissues is a time-consuming task, and the performance depends on the expert’s efficiency. Hence automating tumor segmentation plays a vital role in medical imaging applications. This study aims to provide a comprehensive survey on recent works developed in brain tumor segmentation. In this paper, a systematic literature review is presented to the reader to understand three policies, namely classical scheme, machine learning strategy, and deep learning methodology meant for tumor segmentation. Our primary goal is to include classical methods like atlas-based strategy and statistical-based models employed for segmenting tumors from brain MRI. Few studies that utilized machine learning approaches for the segmentation and classification of brain structures are also discussed. After that, the study provides an overview of deep learning-based segmentation models for quantitative analysis of brain MRI. Deep learning plays a vital role in the automatic segmentation of brain tissues. Presently deep learning technique outshines traditional statistical methods and machine learning approaches. An effort is made to enclose the literature on patch-based and semantic-based tissue segmentation presented by researchers working in the discipline of medical imaging. The manuscript discusses the basic convolutional neural network architecture, Data Sets, and the existing deep learning techniques for tissue segmentation coupled with classification. This paper also attempts to summarize the current works in Convolutional Neural networks and Autoencoders that assist researchers in seeking future directions. Finally, this article is concluded with possible developments and open challenges in brain tumor segmentation.
In industrial manufacturing systems, failures of machines caused by faults in their key components greatly influence operational safety and system reliability. Many data-driven methods have been developed for machinery diagnostics and prognostics. However, there lacks sufficient labeled data to train a high-performance data-driven model. Moreover, machinery datasets are usually collected from different operation conditions and mechanical components, leading to poor model generalization. To address these concerns, cross-domain transfer learning methods are applied to enhance the feasibility and accuracy of data-driven methods for machinery diagnostics and prognostics. This paper presents a comprehensive survey about how recent studies apply diverse transfer learning methods into machinery tasks including diagnostics and prognostics. Three types of commonly-used transfer methods, i.e., model and parameter transfer, feature matching and adversarial adaptation, are systematically summarized and elaborated on their main ideas, typical models and corresponding representative studies on machinery diagnostics and prognostics. In addition, ten widely-used open-source machinery datasets are presented. Based on recent research progress, this survey expounds emerging challenges and future research directions of transfer learning for industrial applications. This survey presents a systematic review of recent research with clear explanations as well as in-depth insights, thereby helping readers better understand transfer learning for machinery diagnostics and prognostics.
Moth flame optimization (MFO) algorithm is a relatively new nature-inspired optimization algorithm based on the moth’s movement towards the moon. Premature convergence and convergence to local optima are the main demerits of the algorithm. To avoid these drawbacks, a modified dynamic opposite learning-based MFO algorithm (m-DMFO) is presented in this paper, incorporating a modified dynamic opposite learning (DOL) strategy. To validate the performance of the proposed m-DMFO algorithm, it is tested via twenty-three benchmark functions, IEEE CEC’2014 test functions and compared with a wide range of optimization algorithms. Moreover, Friedman rank test, Wilcoxon rank test, convergence analysis, and diversity measurement have been conducted to measure the robustness of the proposed m-DMFO algorithm. The numerical results show that, the proposed m-DMFO algorithm achieved superior results in more than 90% occasions. The proposed m-DMFO achieves the best rank in Friedman rank test and Wilcoxon rank test respectively. In addition, four engineering design problems have been solved by the suggested m-DMFO algorithm. According to the results, it achieves extremely impressive results, which also illustrates that the algorithm is qualified in solving real-world problems. Analyses of numerical results, diversity measure, statistical tests and convergence results ensure the enhanced performance of the proposed m-DMFO algorithm.
The flow chart of the proposed method for better understanding
A concept lattice for granulation (0.2, 0.4, 0.8) context shown in Table 10
Recent time characterization of domain based expert of given field is considered as one of the crucial tasks due to uncertainty and randomness in document publication. It becomes more crucial when interdisciplinary, collaborative and other uncertain papers are published by an author or institute only to receive the ranking. In this case precise characterization of founding author or institute of given domain generate uncertainty. This problem starts because document publications, collaboration or expert analysis of given field is totally dark data set which contains lots of unstructured, incomplete or uncertain data. Due to which, less attention has been paid towards this direction. However, it impacts more to recruitment process analysis, brain drain analysis, reviewer comment analysis, collaborative publication analysis, loyal or honest author analysis, or even conflict of interest analysis. To control this issue, a method is proposed in this paper to characterize the the document published by any institute or author in true, false or indeterminant zone of given domain using the properties of single–valued neutrosophic set. The expert of the given domain is classified based on defined (α,β,γ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\alpha, \beta, \gamma $$\end{document})–cut. Same time another method is proposed to deal the case of higher randomness in publication due to multiple author or collaboration using Shannon entropy. The obtained results from both of the methods are compared with each other as well as recently available approaches for validation.
Online education has been facing difficulty in predicting the academic performance of students due to the lack of usage of learning process, summative data and a precise prediction of quantitative relations between variables and achievements. To address these two obstacles, this study develops an artificial intelligence-enabled prediction model for student academic performance based on students’ learning process and summative data. The prediction criteria are first predefined to characterize and convert the learning data in an online engineering course. An evolutionary computation technique is then used to explore the best prediction model for the student academic performance. The model is validated using another online course that applies the same pedagogy and technology. Satisfactory agreements are obtained between the course outputs and model prediction results. The main findings indicate that the dominant variables in academic performance are the knowledge acquisition, the participation in class and the summative performance. The prerequisite knowledge tends not to play a key role in academic performance. Based on the results, pedagogical and analytical implications are provided. The proposed evolutionary computation-enabled prediction method is found to be a viable tool to evaluate the learning performance of students in online courses. Furthermore, the reported genetic programming model provides an acceptable prediction performance compared to other powerful artificial intelligence methods.
This study aims to investigate a general optimization procedure of the Hedge-algebras controller (HAC) for controlling dynamic systems. Based on the analysis of factors affecting the control efficiency of HAC, the optimization problem is established following a multi-objective approach. When optimizing HAC, the design variables contain tuning coefficients of control rules, selections of linguistic terms of each rule in the rule base, fuzziness measure parameters of linguistic variables, and variations of the reference range of state and control variables. In which the proposed tuning coefficients have been improved compared with the previous study. In particular, a new inference method is proposed based on the shape/interpolation function of the finite element method. A three-story building structure subjected to earthquake loads is used in the simulation as a case study to demonstrate the effectiveness of the proposed approach. Research results in the present work show that the proposed procedure is general and can be utilized to control different dynamic systems. Moreover, as mentioned above, a large number of the design variables will cause a significant variation in objective functions. It means that the optimum performance is improved compared to optimal cases of individual design variables.
Number of papers on diabetes subtyping by Scopus per year
Graphical illustration of diabetes subtyping based on (Ahlqvist et al. 2018)
Types of clustering algorithms
Visualization of the number of each clustering algorithms implemented based on previous studies
Healthcare is evolving from standard to personalized, driven by the patients’ needs. Personalized healthcare is a medical model based on genetics, genomics, and other biological information that helps to predict risk for disease. To date, machine learning and data mining are the fastest-growing healthcare field used to classify patient cohorts from a large dataset and its application for diabetes subtyping will be a breakthrough. In this review paper, we have identified, analyzed, and summarized how previous studies distinguished diabetes into subtypes besides implementing the methods for diabetes subtyping using data mining and various clustering algorithms. We have discovered that many studies have suggested diabetes can be differentiated into subtypes clinically based on the risk complications, genetically defined, using clinical features, and for treatment selection. As for clustering algorithms, k-means clustering and hierarchical clustering were shown to be widely used in determining sub-clusters of diabetes. To further investigate diabetes subtyping, understanding the specific objective and method of diabetes subtyping using clustering algorithms from a large dataset will be crucial which could contribute to novel knowledge and improvement for diabetes management.
Face recognition has long been an active research area in the field of artificial intelligence, particularly since the rise of deep learning in recent years. In some practical situations, each identity has only a single sample available for training. Face recognition under this situation is referred to as single sample face recognition and poses significant challenges to the effective training of deep models. Therefore, in recent years, researchers have attempted to unleash more potential of deep learning and improve the model recognition performance in the single sample situation. While several comprehensive surveys have been conducted on traditional single sample face recognition approaches, emerging deep learning based methods are rarely involved in these reviews. Accordingly, we focus on the deep learning-based methods in this paper, classifying them into virtual sample methods and generic learning methods. In the former category, virtual images or virtual features are generated to benefit the training of the deep model. In the latter one, additional multi-sample generic sets are used. There are three types of generic learning methods: combining traditional methods and deep features, improving the loss function, and improving network structure, all of which are covered in our analysis. Moreover, we review face datasets that have been commonly used for evaluating single sample face recognition models and go on to compare the results of different types of models. Additionally, we discuss problems with existing single sample face recognition methods, including identity information preservation in virtual sample methods, domain adaption in generic learning methods. Furthermore, we regard developing unsupervised methods is a promising future direction, and point out that the semantic gap as an important issue that needs to be further considered.
Feature selection is one basic and critical technology for data mining, especially in current “big data era”. Rough set theory is sensitive to noise in feature selection due the stringent condition of an equivalence relation. However, D–S evidence theory is flexible to measure uncertainty of information. In this paper, we introduce robust feature evaluation metrics “belief function” and “plausibility function” into feature selection algorithm to avoid the defect that classification effect is affected by noise such as missing values, confusing data, etc. Firstly, similarity between information values in a set-valued information system (SVIS) is introduced and a variable parameter to control the similarity of samples is given. Secondly, θ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta$$\end{document}-lower and θ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta$$\end{document}-upper approximations in an SVIS are put forward. Then, the concepts of θ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta$$\end{document}-belief function, θ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta$$\end{document}-plausibility function, θ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta$$\end{document}-belief reduction and θ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\theta$$\end{document}-plausibility reduction are given. Moreover, several feature selection algorithms based on the D–S evidence theory in an SVIS are proposed. Experimental results and statistical test show that the proposed metric is insensitive to noise because it comprehensively considers the evidence at all levels, and the proposed algorithms are more robust than several state-of-the-art feature selection algorithms.
Case-Based Reasoning (CBR) is an artificial intelligence approach to problem-solving with a good record of success. This article proposes using Quantum Computing to improve some of the key processes of CBR, such that a quantum case-based reasoning (qCBR) paradigm can be defined. The focus is set on designing and implementing a qCBR based on the variational principle that improves its classical counterpart in terms of average accuracy, scalability and tolerance to overlapping. A comparative study of the proposed qCBR with a classic CBR is performed for the case of the social workers’ problem as a sample of a combinatorial optimization problem with overlapping. The algorithm’s quantum feasibility is modelled with docplex and tested on IBMQ computers, and experimented on the Qibo framework.
Learning from demonstration, or imitation learning, is the process of learning to act in an environment from examples provided by a teacher. Inverse reinforcement learning (IRL) is a specific form of learning from demonstration that attempts to estimate the reward function of a Markov decision process from examples provided by the teacher. The reward function is often considered the most succinct description of a task. In simple applications, the reward function may be known or easily derived from properties of the system and hard coded into the learning process. However, in complex applications, this may not be possible, and it may be easier to learn the reward function by observing the actions of the teacher. This paper provides a comprehensive survey of the literature on IRL. This survey outlines the differences between IRL and two similar methods - apprenticeship learning and inverse optimal control. Further, this survey organizes the IRL literature based on the principal method, describes applications of IRL algorithms, and provides areas of future research.
With the development of Computer-aided Diagnosis (CAD) and image scanning techniques, Whole-slide Image (WSI) scanners are widely used in the field of pathological diagnosis. Therefore, WSI analysis has become the key to modern digital histopathology. Since 2004, WSI has been used widely in CAD. Since machine vision methods are usually based on semi-automatic or fully automatic computer algorithms, they are highly efficient and labor-saving. The combination of WSI and CAD technologies for segmentation, classification, and detection helps histopathologists to obtain more stable and quantitative results with minimum labor costs and improved diagnosis objectivity. This paper reviews the methods of WSI analysis based on machine learning. Firstly, the development status of WSI and CAD methods are introduced. Secondly, we discuss publicly available WSI datasets and evaluation metrics for segmentation, classification, and detection tasks. Then, the latest development of machine learning techniques in WSI segmentation, classification, and detection are reviewed. Finally, the existing methods are studied, and the application prospects of the methods in this field are forecasted.
Top-cited authors
Jim Austin
  • The University of York
Victoria Hodge
  • The University of York
Asifullah Khan
  • Pakistan Institute of Engineering and Applied Sciences
Anabia Sohail
  • Air University of Islamabad
Umme Zahoora
  • Pakistan Institute of Engineering and Applied Sciences