Search

• Robust functional form contains true parameters far more often than popular models. • Matches/outperforms widely used regression and Neural Network models. • Finds appropriate balance between Model Fit, Inference and Prediction (MIPs). • Introduces new large-sample DGP test; can use to improve A.I. models. • For MIS field finds Popularity Parameter to be important for predicting citations. This article introduces a versatile functional form for Generalized Linear Models (GLMs) through a simple, yet effective, transformation of the current framework. The models are applied through a new hierarchical bayesian estimation procedure for logistic regression to highly-cited papers in the Management Information Systems (MIS) field. The results are uniformly better, in regards to model fit and inference for in-sample and out-of-sample data, for simulation studies and real-world data applications, requiring very little time to convergence to true population parameters. In simulation studies, I show that the method contains the true parameters nearly three times as often as widely used existing GLMs, and does so while having confidence intervals that are 54.50% smaller, while requiring around two-thirds the number of MCMC iterations as existing bayesian methods. In Scientometric applications the methodology is shown to be highly robust with predictive/classification accuracy, either equaling or exceeding existing methods for identifying highly-cited articles including Artificial Neural Networks (ANN). Thus, the method is shown to be robust to the amount of asymmetry (or symmetry) of the probability of success (or failure) and robust to unbalanced samples and varying Data Generating Processes. Further, the methodology is equivalent to current methods if the data support them and is therefore complementary to existing methods, without loss of interpretability of model parameters. For the MIS field it finds that Popularity Parameter (PP) of an article Keywords can predict whether a paper will be highly-cited (top 25% of highly-cited articles) between two to three years after publication and beyond. Furthermore, given the small number of iterations needed for convergence, the methodology can also be used as a baseline method in Big Data (BD) settings for both Artificial Intelligence (AI) and Machine Learning (ML) contexts as well. [ABSTRACT FROM AUTHOR]

Malaria is a significant health concern worldwide, particularly in Africa where its prevalence is still alarmingly high. Using artificial intelligence algorithms to diagnose cells with malaria provides great convenience for clinicians. In this paper, a densely connected convolutional dynamic learning network (DCDLN) is proposed for the diagnosis of malaria disease. Specifically, after data processing and partitioning of the dataset, the densely connected block is trained as a feature extractor. To classify the features extracted by the feature extractor, a classifier based on a dynamic learning network is proposed in this paper. Based on experimental results, the proposed DCDLN method demonstrates a diagnostic accuracy rate of 97.23%, surpassing the diagnostic performance than existing advanced methods on an open malaria cell dataset. This accurate diagnostic effect provides convincing evidence for clinicians to make a correct diagnosis. In addition, to validate the superiority and generalization capability of the DCDLN algorithm, we also applied the algorithm to the skin cancer and garbage classification datasets. DCDLN achieved good results on these datasets as well, demonstrating that the DCDLN algorithm possesses superiority and strong generalization performance. [ABSTRACT FROM AUTHOR]

The COVID-19 pandemic exposed the importance of research on the spread of epidemic diseases. In this paper, we apply Artificial Intelligence and statistics techniques to build prediction models to estimate the SARS-CoV-2 seroprevalence in the United States, using multiple estimates of COVID-19 prevalence and other explanatory variables. We propose the use of stacking techniques based on multiple model building techniques (Linear and Beta Regression, Genetic Programming and Neural Networks) to obtain Predictive Ensemble Models. There has been extensive research on this field, but there has not been in-depth research on the application of stacking methods to estimate and forecast seroprevalence in the USA specifically. This paper provides a novel comparison of the behaviour and performance of different building techniques for stacking ensemble models and presents which methods are better for different scenarios. We find that Genetic Programming and Neural Networks are the best models with trained data within single states, and when multiple states are considered Genetic Programming is still better than the Regression models, but Neural Networks fail to estimate the seroprevalence accurately. Another novelty of our work is the use of cross-state validation to evaluate the models with new data, as well as temporal forecasting. Depending on how the data is processed, Linear Regression performs very well with cross-state validation and temporal forecasting, and Genetic Programming is very accurate with the former while Neural Networks work better with the latter. • COVID-19 seropositivity modelling in the USA. • Use of stacking and ensemble ML models. • Comparison of ML methods (linear regression, beta regression, genetic programming, neural network). • Trade-off between explainability and performance. [ABSTRACT FROM AUTHOR]

In the quest for sustainable manufacturing, a critical yet often overlooked aspect is predicting the scrap in nesting-based manufacturing. This aspect is pivotal for its eco-friendly implications and potential to enhance cost-efficiency and competitiveness in the manufacturing industry. This paper introduces a method to predict scrap generation in nesting-based manufacturing solely based on part geometries without prior knowledge of the resulting arrangement. Our method employs three deep learning modules: one for encoding geometries, another to condense these into an information vector reflecting interdependence for scrap prediction, and a third for the actual prediction using the prior modules. Evaluated on a private real-world sheet metal dataset, the technique achieves a Mean Absolute Percentage Error (MAPE) of 24.8%, showcasing its precision in scrap prediction. Such predictions can bolster sustainable manufacturing by curtailing material waste, improving production efficiency, and enhancing eco-sustainability. The paper underscores scrap prediction's role in advancing sustainable manufacturing in nesting-based industries. [ABSTRACT FROM AUTHOR]

Underwater spectral imaging combined with neural networks provides a practical means for intelligent detection and maintenance of offshore engineering. As an underwater detection technology, the data obtained from underwater hyperspectral imaging can be used for both qualitative analysis and quantitative detection of underwater targets, which is an efficient means of underwater target detection in offshore engineering. In this paper, An underwater hyperspectral imaging system based on PGP spectral structure line scanning imaging has been designed and developed for shallow water detection. It has a spectral range of 400–1000 nm, over 200 channels, a spectral resolution of 2.5 nm, a field of view angle of 35.2°, and a focal length of 25 mm. According to the design results, the system was assembled, calibrated and tested. Aiming at solving the shortcomings of the current underwater hyperspectral system, such as the inability of real-time observation of the scanned image, large data volume and complicated processing, the upper computer software was designed and developed to realize the functions of controlling and adjusting the parameter, real-time displaying the scanned image, reconstructing the image and outputting the data in ENVI standard format to the underwater hyperspectral system. The prototype of the system was used to conduct coral detection experiments in Weizhou Island, and the neural network model was used to classify the experimental data, and the overall classification accuracy was above 98%. The experimental results show that the whole system has good imaging quality and good underwater detection capability, and the developed host computer software greatly reduces the workload of raw data processing, improves the experimental detection efficiency, and provides a feasible technical solution for underwater intelligent detection and analysis. • A line-scanning underwater hyperspectral imaging system has been developed based on the PGP spectroscopic structure. • Development of software for the upper computer, which significantly improves the efficiency of underwater detection. • Combining neural network models to achieve accurate detection and classification of underwater targets. [ABSTRACT FROM AUTHOR]

Artificial intelligence models encounter significant challenges due to their black-box nature, particularly in safety-critical domains such as healthcare, finance, and autonomous vehicles. Explainable Artificial Intelligence (XAI) addresses these challenges by providing explanations for how these models make decisions and predictions, ensuring transparency, accountability, and fairness. Existing studies have examined the fundamental concepts of XAI, its general principles, and the scope of XAI techniques. However, there remains a gap in the literature as there are no comprehensive reviews that delve into the detailed mathematical representations, design methodologies of XAI models, and other associated aspects. This paper provides a comprehensive literature review encompassing common terminologies and definitions, the need for XAI, beneficiaries of XAI, a taxonomy of XAI methods, and the application of XAI methods in different application areas. The survey is aimed at XAI researchers, XAI practitioners, AI model developers, and XAI beneficiaries who are interested in enhancing the trustworthiness, transparency, accountability, and fairness of their AI models. [ABSTRACT FROM AUTHOR]

The commercialization of Fuel Cell Systems (FCSs) faces a key obstacle in the form of limited cell lifespan and resultant cell degradation. This paper introduces a new approach, leveraging an artificial intelligence-based 2DOF control system to regulate the Oxygen Excess Ratio (OER) with two main goals: mitigating cell degradation due to oxygen starvation and optimizing net power output. The proposed control system incorporates a data-driven feedforward controller in conjunction with a feedback controller, facilitating the tracking of desired OER values generated by a data-driven reference generator. We develop fuzzy models and neural networks as the reference generator and feedforward controller to capture the complex FCS behavior by processing the stack current w/wo the temperature of FCS (i.e. Single input or Double input models). By exploring the effects of the structural settings of the models, this study provides a comprehensive understanding of their impact on the representation performance of the FCS characteristics. Although the fitting performances of all models are quite satisfactory, their actual performance gain is evaluated on a realistic FCS model at various operation points. The findings and comparative analysis emphasize the efficacy of incorporating stack temperature in fuzzy-model-based 2DOF control systems, showcasing their potential to maximize net power output through enhanced OER control loop performance while also extending the lifespan of FCSs, as confirmed by results from a developed degradation model. • Improved net output power through the integration of AI in a 2DOF control system. • Reduction of degradation using a data-driven 2DOF control system. • Development of data-driven reference generators and feedforward controllers. • Comparative performance evaluations of fuzzy models and neural networks. [ABSTRACT FROM AUTHOR]

Modern cities dynamically face several challenges including digitalization, sustainability, resilience and economic development. Urban planners and designers must develop urban forms that address these challenges. With the integration of new communication and information technologies (Smartphone, GIS, Drones, IoT, Sensors, etc.), urban activities have generated large volumes of urban data. The rapid growth in terms of collection and big data storage capacities combined with the ever-increasing computational power of modern machines have made possible their efficient treatment using machine (ML) and deep learning (DL) algorithms. The emergence of such groundbreaking methods has in turn helped to address the challenges of modern-day cities in several domains (health, security, mobility, etc). ML algorithms have been proposed to model the urban form's indicators for intelligent urban planning decision making. They have been proven to perform better than the traditional methods. However, the potential of ML has not yet been fully explored in research for urban planning decision support. This paper presents a comprehensive review of ML applications for mitigating the challenges of modern cities planning. First and foremost, an overview of the urban forms, sources of urban data, the ML and DL techniques as well as their potential in solving the aforementioned challenges. For each ML method, we will highlight it working principle, advantages, disadvantages and potential applications using comparative tables. Finally, we will discuss the issues and challenges of ML methods in urban form's modeling while ultimately advocating some future research directions. [ABSTRACT FROM AUTHOR]

The rising overheads of data-movement and limitations of general-purpose processing architectures have led to a huge surge in the interest in "processing-in-memory" (PIM) approach and "neural networks" (NN) architectures. Spintronic memories facilitate efficient implementation of PIM approach and NN accelerators, and offer several advantages over conventional memories. In this paper, we present a survey of spintronic-architectures for PIM and NNs. We organize the works based on main attributes to underscore their similarities and differences. This paper will be useful for researchers in the area of artificial intelligence, hardware architecture, chip design and memory system. [ABSTRACT FROM AUTHOR]

• If the AI methods properly be developed, they can successfully be used to simulate and forecast the GWL time series. • Several new ideas in the related area of research are presented in this study for developing innovative methods and for improving the quality of the modeling. • The AI methods can be linked to physically/conceptually based models such as MODFLOW to develop integrated modular models such that each method covers the weak points of the other method. • Regarding different AI methods to simulate the GWL, it can be said that it is not practically possible to recommend one particular type of AI model for a given problem. However it is clear that a hybrid/coupled model likely perform better than a single AI model. This study is a review to the special issue on artificial intelligence (AI) methods for groundwater level (GWL) modeling and forecasting, and presents a brief overview of the most popular AI techniques, along with the bibliographic reviews of the experiences of the authors over past years, and the reviewing and comparison of the obtained results. Accordingly, 67 journal papers published from 2001 to 2018 were reviewed in the terms of the features and abilities of the modeling approaches, input data consideration, prediction time steps, data division, etc. From the reviewed papers it can be concluded that despite some weaknesses, if the AI methods properly be developed, they can successfully be used to simulate and forecast the GWL time series in different aquifers. Since some of the stages of the AI modeling are based on the experience or trial-and-error procedures, it is useful to review them in the special application on GWL modeling. Many partial and general results were achieved from the reviewed papers, which can provide applicable guidelines for researchers who want to perform similar works in this field. Several new ideas in the related area of research are also presented in this study for developing innovative methods and for improving the quality of the modeling. [ABSTRACT FROM AUTHOR]

Modern manufacturing environments strive to support a great variety of products, driven by the large customization demand, while at the same time maintain low cost and fast response to the market. To cope with the challenging manufacturing requirements, organizations rely on agile decision-making systems to maintain a smooth production function. In the paint shop environment, where hanging positions are included on a circular conveyor belt for different type of items to be transferred through the painting cabins, the provision of a near optimal scheduling solution is a rather important and complex aspect due to the diversity in colors, sizes, deadlines and more. This research paper aims to address this issue with the development of a toolbox of autonomous agents that provide near optimum solutions in the paint shop scheduling problem, by optimizing the sequence and combination of items transferred through the paint-shop conveyor. The proposed approach considers sequence-based setup time dependencies and capacity constrains for the different item types, while aiming for maximum conveyor utilization and minimum production flowtime. The smart agents' toolbox is empowered with both model-based and data-driven optimization methods allowing it to avoid long computational delays by exploiting the different benefits of each scheduling agent. The agent interface was implemented in Java, whereas Python was used for mathematical modeling, analysis, and optimization. The research validation was performed based on real industrial data provided by a bicycle industry where the agents were tested in scheduling the painting department. Results from the performance of the different optimization methodologies within the pilot case, were used to indicate the advantages and disadvantages of each method with respect to the problem's characteristic. [ABSTRACT FROM AUTHOR]

For large-scale terahertz (THz) reconfigurable metamaterials (RM), each unit element needs to be controlled by an independent voltage source. This leads to a huge solution space for control words to achieve the required RM performance specification. In this paper, we propose a fast AI-assisted control-word generation scheme to reduce the computation cost of electromagnetic (EM) simulations (Forward Model) and to speed up iterations for control word searches (Inverse Model). We demonstrate the effectiveness of our models by designing and fabricating a 12 × 9 RM array with a Bi-layer Bi-Split-Ring Resonator structure (B-BSRR) in standard 180 nm CMOS technology. The terahertz RM can be electrically reconfigured, providing active metamaterials with an amplitude modulation depth of around 0–1.7 dB at 0.49 THz. The results show that our Forward Model can predict the S-parameters between 100 GHz and 800 GHz with a mean absolute error of 0.61 dB. Our method is more than 200 times faster than traditional full-wave simulation methods. The Inverse Model can generate required control words within a 1.3 dB error margin in less than 200 iterations. • Using graph neural network and deep learning to construct a neural network that can predict S21. • Proposed an inverse model to accelerate the generation of control words. • Completed a prototype test chip in 180 nm CMOS technology with amplitude modulation depth up to 0.85 dB. [ABSTRACT FROM AUTHOR]

This paper mainly deals with the problem of designing delayed state-feedback controller for neural networks with leakage delay. By constructing an appropriate Lyapunov–Krasovskii functional including double integral terms having two different exponential decay rates and utilizing linear matrix inequality (LMI) technique with the help of slack variables, some sufficient conditions for globally exponential stabilization results are obtained by designing of delayed state-feedback controller. The novelty of this paper includes: (i) although many papers dealt with dynamics of neural networks with leakage delay, there is little work on design of feedback controller. Even there is almost no result that deals with the delay-feedback controller for such delay systems, which is the main motivation of this paper; (ii) the derived delay-dependent stability criteria establish the relationship between leakage delay and time delay in the feedback term, which can be easily checked via the MATLAB LMI toolbox; (iii) we consider the general case that the control term is given in the form of B u , where B is an n × m real matrix. Such case is more difficult to handle than the special case that B is unit matrix. It indicates that the controller can be used in all states or in some states by proper selection of the matrix B . The effectiveness of the proposed method in this paper is illustrated via numerical examples. [ABSTRACT FROM AUTHOR]

• The DRL employs the Using Graph Convolutional Network (GCN) to extract the network architecture data, the node vector is created by fusing it with traffic-related features. • The deep Q-network uses this node vector to anticipate the best rewards and determine which FDNN task is carried out at the edge nodes. • In light of this choice, the FDNN classifier is deployed to categorize the medical information. • In the end, the simulation findings show that the DRL-LOA significantly enhances optimization, showcasing a 35.7% reduction in power usage and a of 27.9% decrease in latency on 20 edge nodes in cloud-edge systems. • These findings underscore DRL-LOA's efficacy in optimizing medical data categorization and addressing load-balancing challenges in cooperative cloud-edge networks. Addressing the limitations of traffic-centric approaches in cooperative cloud-edge networks, this paper introduces an adaptive deployment strategy for FDNN using LOA. However, the oversight of network structure hampers load-balancing efficiency in medical data categorization. To overcome this, a novel DRL-LOA approach integrates Deep Reinforcement Learning (DRL) with LOA, considering both network structure and traffic-related data. This DRL-LOA considers both network structure and traffic-related data for optimization tasks. The DRL employs the Using Graph Convolutional Network (GCN) to extract the network architecture data, the node vector is created by fusing it with traffic-related features. The deep Q-network uses this node vector to anticipate the best rewards and determine which FDNN task is carried out at the edge nodes. In light of this choice, the FDNN classifier is deployed to categorize the medical information. In the end, the simulation findings show that the DRL-LOA significantly enhances optimization, showcasing a 35.7% reduction in power usage and a of 27.9% % decrease in latency on 20 edge nodes in cloud-edge systems. These findings underscore DRL-LOA's efficacy in optimizing medical data categorization and addressing load-balancing challenges in cooperative cloud-edge networks. [ABSTRACT FROM AUTHOR]

• Low-cost measurement platform for particulate matter monitoring developed. • Artificial intelligence field calibration procedure of low-cost sensor proposed. • Combination of affine scaling and optimized neural network surrogate employed. • Correlation coefficient improved to 0.86 for particulate matter of size 1 μm. • Root-mean-squared error of 3 μm/m3 (1 μm particles) and 4.9 μm/m3 (10 μm particles) Due to detrimental effects of atmospheric particulate matter (PM), its accurate monitoring is of paramount importance, especially in densely populated urban areas. However, precise measurement of PM levels requires expensive and sophisticated equipment. Although low-cost alternatives are gaining popularity, their reliability is questionable, attributed to sensitivity to environmental conditions, inherent instability, and manufacturing imperfections. The objectives of this paper include (i) introduction of an innovative approach to field calibration for low-cost PM sensors using artificial intelligence methods, (ii) implementation of the calibration procedure involving optimized artificial neural network (ANN) and combined multiplicative and additive correction of the low-cost sensor readings, (iii) demonstrating the efficacy of the presented technique using a custom-designed portable PM monitoring platform and reference data acquired from public measurement stations. The results obtained through comprehensive experiments conducted using the aforementioned low-cost sensor and reference data demonstrate remarkable accuracy for the calibrated sensor, with correlation coefficients of 0.86 for PM 1 and PM 2.5 , and 0.76 PM 10 (particles categorized as having diameter equal to or less than 1 μm, 2.5 μm, and 10 μm, respectively), along with low RMSE values of only 3.1, 4.1, and 4.9 µg/m3. [ABSTRACT FROM AUTHOR]

Background and objective: This paper focuses on nutritional recommendation systems (RS), i.e. AI-powered automatic systems providing users with suggestions about what to eat to pursue their weight/body shape goals. A trade-off among (potentially) conflictual requirements must be taken into account when designing these kinds of systems, there including: (i) adherence to experts' prescriptions, (ii) adherence to users' tastes and preferences, (iii) explainability of the whole recommendation process. Accordingly, in this paper we propose a novel approach to the engineering of nutritional RS, combining machine learning and symbolic knowledge extraction to profile users—hence harmonising the aforementioned requirements. Methods Our contribution focuses on the data processing workflow. Stemming from neural networks (NN) trained to predict user preferences, we use CART Breiman et al.(1984) to extract symbolic rules in Prolog Körner et al.(2022) form, and we combine them with expert prescriptions brought in similar form. We can then query the resulting symbolic knowledge base via logic solvers, to draw explainable recommendations. Results Experiments are performed involving a publicly available dataset of 45,723 recipes, plus 12 synthetic datasets about as many imaginary users, and 6 experts' prescriptions. Fully-connected 4-layered NN are trained on those datasets, reaching ∼ 86 % test-set accuracy, on average. Extracted rules, in turn, have ∼ 80 % fidelity w.r.t. those NN. The resulting recommendation system has a test-set precision of ∼ 74 %. The symbolic approach makes it possible to devise how the system draws recommendations. Conclusions Thanks to our approach, intelligent agents may learn users' preferences from data, convert them into symbolic form, and extend them with experts' goal-directed prescriptions. The resulting recommendations are then simultaneously acceptable for the end user and adequate under a nutritional perspective, while the whole process of recommendation generation is made explainable. [ABSTRACT FROM AUTHOR]

Nonparametric cost estimation in construction projects with the use of artificial networks is presented as suitable mainly for the early estimates. These conceptual estimates are based on the variables – namely cost predictors that characterize the project or a facility. Data gathered on the basis of completed projects are combined together and applied to the current project cost estimation process. The aim of the paper is to discuss the opportunities and challenges of the approach based on the artificial intelligence tools to cost estimation of construction works. The proposed approach is based on the concept of nonparametric cost estimation and application of artificial neural networks. The author's idea and intention is to transfer the mechanisms of nonparametric cost estimating to the level of construction works. Neural networks, due to their general capabilities, seem to be a good tool to aid the proposed approach. The paper contains discussion of the proposed approach and its applicability. [ABSTRACT FROM AUTHOR]

The correct and consistent identification of the petrographic properties of coal is an important issue for researchers in the fields of mining and geology. As part of the study described in this paper, investigations concerning the application of artificial intelligence methods for the identification of the aforementioned characteristics were carried out. The methods in question were used to identify the maceral groups of coal, i.e. vitrinite, inertinite, and liptinite. Additionally, an attempt was made to identify some non-organic minerals. The analyses were performed using pattern recognition techniques (NN, kNN), as well as artificial neural network techniques (a multilayer perceptron – MLP). The classification process was carried out using microscopy images of polished sections of coals. A multidimensional feature space was defined, which made it possible to classify the discussed structures automatically, based on the methods of pattern recognition and algorithms of the artificial neural networks. Also, from the study we assessed the impact of the parameters for which the applied methods proved effective upon the final outcome of the classification procedure. The result of the analyses was a high percentage (over 97%) of correct classifications of maceral groups and mineral components. The paper discusses also an attempt to analyze particular macerals of the inertinite group. It was demonstrated that using artificial neural networks to this end makes it possible to classify the macerals properly in over 91% of cases. Thus, it was proved that artificial intelligence methods can be successfully applied for the identification of selected petrographic features of coal. [ABSTRACT FROM AUTHOR]

The paper presents the results regarding the identification of the value of the pull-off adhesion between a concrete added layer with a constant thickness and a substrate concrete layer in existing elements. A method of identification, which is based solely on the use of artificial neural networks (ANNs) and two non-destructive acoustic methods: impact-echo and impulse response on the surface of the added layer, was developed. The methodology of identifying the pull-off adhesion between a concrete added layer and a substrate layer in existing elements was developed and presented in the paper and is useful in construction practice. [ABSTRACT FROM AUTHOR]

Shorter product life cycles and increasing individualization of production leads to an increased reconfiguration demand in the domain of industrial automation systems, which will be dominated by cyber-physical production systems in the future. In constantly changing systems, however, not all configuration alternatives of the almost infinite state space are fully understood. Thus, certain configurations can lead to process instability, a reduction in quality or machine failures. Therefore, this paper presents an approach that enhances an intelligent Digital Twin with a self-organized reconfiguration management based on adaptive process models in order to find optimized configurations more comprehensively. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) technology has significant potential in Earth sciences, particularly in mineral identification for industrial exploration, geological mapping, and archaeological research. However, traditional methods are time-consuming, expensive, and complex. And existing mineral identification methods based on mineral photos face several critical challenges, including lack of consideration for natural image features captured in real environments, limitations of single-label classification which does not align with multi-mineral occurrences in nature, and growing computational complexity as the number of identifiable mineral labels increases. Therefore, this paper proposes an efficient mineral identification model based on multi-label image classification, focusing on natural environmental features. First, realistic feature datasets are created by simulating mineral photos in real environments. Then, the model uses the query-label (Query2Label) framework, with MaxViT-T (Multi-Axis Vision Transformer-Tiny) as the feature extraction network and the asymmetric loss function. Knowledge distillation is employed to improve identification accuracy while reducing computational complexity. The proposed model achieves an impressive average identification accuracy of 84.74% on a dataset of 495,756 mineral photos, surpassing existing models like ResNet-101, ML-GCN (Multi-Label Graph Convolutional Network), and SRN (Spatial Regularization Net). It maintains a lower parameter count and computational complexity. In the end, ablation experiments demonstrate the effectiveness of each optimization scheme. • Selecting image noise and color shift methods for realistic mineral data simulation. • Proposing an enhanced two-stage image multi-Label identification model framework. • Applying knowledge distillation for multi-label classification subtask decomposition. • Being with low computational complexity and high mineral identification accuracy. [ABSTRACT FROM AUTHOR]

• The artificial intelligence approach for modeling adsorption processes in adsorption cooling and desalination systems with fluidized adsorption beds is presented. • The prediction performance of four predictive methods is compared. • The proposed model shows forecast skill score in a range of about 97%–99%. • GRU method has high accuracy and shortest computation time. The article presents deep learning methods applied to predict the mass of an adsorption bed in the fixed and fluidized bed. The purpose of the application of this kind of bed is to improve the efficiency of the adsorption cooling systems by increased heat and mass transfer by using fluidization. The paper employs three deep learning methods: Long Short-Term Memory (LSTM), Bidirectional Long Short-Term Memory (BiLSTM), and Gated Recurrent Unit (GRU). In each of the selected neural networks, the epoch values, the number of neurons on the first layer and the number of neurons on the second layer were changed. These networks had two hidden layers. The paper presents numerical research on mass prediction using the algorithm mentioned above for silica gel as sorbent with copper, aluminum, carbon nanotubes additives. The results obtained by the developed algorithms of the LSTM, BiLSTM, GRU network and experimental tests are in good agreement with R 2 above 0.97. The GRU network guarantees predicting the mass of both the fluidized and fixed beds with the best agreement with the measurement results. [ABSTRACT FROM AUTHOR]

The efficiency of solar tower power plants depends strongly on the ability to reflect the sun light onto a defined point on the receiver. Due to the high demands on the heliostats to achieve high accuracy at low costs, a regular calibration is necessary to reduce the tracking error. In this paper a new method for improving existing calibration methods using deep learning is presented. The results are validated by using calibration data recorded at the Solar Tower Jülich with the example of one heliostat. Through a combination of Self-Normalizing Neural Networks and transfer learning it is possible to benefit from the advantages of neural networks already with a training dataset of only 300 measuring points. With that a measured test accuracy of 0.42 mrad was achieved. This was approximately three times more accurate than the best result of the compared state-of-the-art regression algorithm used in Jölich. Furthermore we give recommendations on the structure of the dataset and the neural network (NN) pretraining necessary for these results. [ABSTRACT FROM AUTHOR]

New technologies are transforming medicine, and this revolution starts with data. Health data, clinical images, genome sequences, data on prescribed therapies and results obtained, data that each of us has helped to create. Although the first uses of artificial intelligence (AI) in medicine date back to the 1980s, it is only with the beginning of the new millennium that there has been an explosion of interest in this sector worldwide. We are therefore witnessing the exponential growth of health-related information with the result that traditional analysis techniques are not suitable for satisfactorily management of this vast amount of data. AI applications (especially Deep Learning), on the other hand, are naturally predisposed to cope with this explosion of data, as they always work better as the amount of training data increases, a phase necessary to build the optimal neural network for a given clinical problem. This paper proposes a comprehensive and in-depth study of Deep Learning methodologies and applications in medicine. An in-depth analysis of the literature is presented; how, where and why Deep Learning models are applied in medicine are discussed and reviewed. Finally, current challenges and future research directions are outlined and analysed. • We review the state-of-the-art focusing on the application of DL in medicine. • We expose a categorization of Deep Learning models used and applied in medicine. • We classify medicine-related DL applications into macro-areas and sub-areas. • We thoroughly discuss the recent and open challenges related to DL in medicine. [ABSTRACT FROM AUTHOR]

Predictive maintenance (PdM) for smart manufacturing and Industry 4.0 has been associated with manufacturing intelligence supported by Artificial Intelligence (AI). Therefore, PdM also relies on the smart manufacturing technologies including cyber-physical system (CPS) and big data analytics. The multi-agent system (MAS) technology and deep learning (DL) have shown the capacity to provide efficient tools for the implementation of PdM in a CPS enabled smart industrial production system gaining feedback from big data analytics. Induction motors (IM) constitute the main power source in the industrial production environment and therefore their maintenance and early fault detection and diagnosis (FD/D) is a critical process. Neural network (NN) based FD/D of IM has been widely used in order to identify different fault types. DL methods have recently emerged for FD/D of IM and can efficiently analyze massive data coming from different machine sensors. The MAS has recently been used in combination with artificial NNs as a decision support tool for FD/D of IM. This paper aims to provide a review of recent trends in PdM of IM focusing on MAS and DL based FD/D methods that have emerged in the last 5 years due to their potential to be implemented in a smart manufacturing system. A discussion of the presented methods is given in order to present the recent developments and trends and provide future directions for research. [ABSTRACT FROM AUTHOR]

While high-efficiency propulsion techniques are enabling new mission concepts in deep space exploration, their limited thrust capabilities necessitate long thrusting arcs and make spacecraft more susceptible to missed thrust events. To correct for such mishaps, most spacecraft require updated trajectories that are relayed from Earth. While this solution is viable for spacecraft near Earth, in deep space, where one-way communication time is measured in hours, a delay in transmission may prolong the time of flight or result in a complete loss of mission. Such problems can be alleviated by increasing the spacecraft's onboard autonomy in guidance. This paper demonstrates how a computationally lightweight neural network can map the spacecraft's state to a near-optimal control action, autonomously guiding a spacecraft within different astrodynamic regimes and optimality criteria. The neural network is trained using supervised learning and datasets comprised of optimal state-action pairs, as determined through traditional direct and indirect methods. Additionally, the neural network-designed solutions retain optimality and time of flight corresponding to traditional trajectories. Finally, the same neural networks can autonomously correct for most missed thrust events encountered on long-duration low-thrust trajectories. The presented results provide a path for mitigating risks associated with the use of high-efficiency low-thrust propulsion techniques. • Deep neural networks can autonomously guide low-thrust spacecraft. • The final costs are on average, less than 3% different the optimal trajectories. • Can autonomously correct for >75% of missed thrust events encountered. [ABSTRACT FROM AUTHOR]

Currently, there is a need in society to use digital and cybernetic methods of remote diagnostics of a person's mental state. Using a digital or cyberphysical system in the process of remote recognition of the patient's mental state, it is possible to increase the effectiveness of recognizing the actual psycho-emotional state of the patient, both in the consulting mode and in dynamic observation. For remote assessment of the mental state, methods are widely used based on recognition of facial activity, which is reflexively associated with the display of the psycho-emotional state. In this paper, the authors propose a new approach to creating such a methodology. This work was supported by the RFBR grant 18-07-00987. [ABSTRACT FROM AUTHOR]

Due to industrialization and the growth of transportation systems, the number of vehicles continues to increase which causes a significant traffic jam problem especially in big cities. Consequently, the prediction of traffic flows is a key component to an optimal traffic management. As a solution to this issue, the present paper aims at applying the artificial intelligence of neural networks, which offers an interesting approach to modelling in complex, and nonlinear situations. Our resolution method is based on the design of a neural network to predict daily traffic flow. Then, the forecasted traffic flow is compared with a real dataset recorded on a road section and provided by a recognized infrastructure manager in Morocco. Indeed, neural networks have the ability to learn from the past and predict the future. In this study, various neural networks structures are exanimated and simulation results show that the best forecasts are obtained with the use of Multi-Layer Perceptron architecture that has a good generalization capacity with a total Mean Square Error of 0.00927 in the train set and 0.01321 in the test set. [ABSTRACT FROM AUTHOR]

Neural networks are widely used in various applications such as image classification, speech recognition and natural language processing. As intelligent systems often rely on neural networks to make critical decisions, it is crucial to ensure that the behavior of a neural network is trustworthy. Along with the increase in the scale and complexity of neural networks, it becomes a challenging problem to verify their safety, since the large-scale state space may make the verification of safety properties infeasible or unsolvable. In order to achieve efficient safety verification of neural networks, this paper proposes a divide-and-conquer verification method based on neuron importance analysis. The method converts the safety constraints to be verified into an optimization problem composed of output neurons, and recursively measures the influence of neurons in the network on the final objective function to calculate their importance. The neurons with the greatest importance are split to refine the constraints, so that the original verification problem is transformed into a set of sub-problems for solution. The method is implemented as a tool named NIAVerify, which is evaluated through experiments of property verification on ReLU-based fully-connected benchmark networks trained on two classic datasets. Experimental results show that NIAVerify can achieve significant state space reduction and the verification efficiency is 44.80% higher than the existing verification method. [ABSTRACT FROM AUTHOR]

Twelve lead electrocardiogram signals capture unique fingerprints about the body's biological processes and electrical activity of heart muscles. Machine learning and deep learning-based models can learn the embedded patterns in the electrocardiogram to estimate complex metrics such as age and gender that depend on multiple aspects of human physiology. ECG estimated age with respect to the chronological age reflects the overall well-being of the cardiovascular system, with significant positive deviations indicating an aged cardiovascular system and a higher likelihood of cardiovascular mortality. Several conventional, machine learning, and deep learning-based methods have been proposed to estimate age from electronic health records, health surveys, and ECG data. This manuscript comprehensively reviews the methodologies proposed for ECG-based age and gender estimation over the last decade. Specifically, the review highlights that elevated ECG age is associated with atherosclerotic cardiovascular disease, abnormal peripheral endothelial dysfunction, and high mortality, among many other cardiovascular disorders. Furthermore, the survey presents overarching observations and insights across methods for age and gender estimation. This paper also presents several essential methodological improvements and clinical applications of ECG-estimated age and gender to encourage further improvements of the state-of-the-art methodologies. • ECG-based age and gender estimation methods are comprehensively reviewed and analyzed. • Methods are stratified based on the type of methodology adopted and the nature of the study (conventional/deep-learning-based and technical/application). • Tabular summaries and textual descriptions of each method are provided, including core methodology, dataset information, key results, and limitations. • Key commonalities and limitations across the reviewed methods are analyzed and presented. • Essential technical aspects of deep learning methods for ECG-based age and gender estimation that need validation for further progress are discussed. • Valuable insights and recommendations for future research and progress are shared. [ABSTRACT FROM AUTHOR]

In the author's previous paper published under the same title, the author showed that in a water-surface wave train, the peak height of a next-coming wave can be forecasted fairly well with recurrent neural networks (RNN). It has been also shown that the motion peak heights of a floating body in a wave train can be forecasted even better than the forecasting of wave peak heights. As the Part-2 paper, freak events such as extremely large waves or motions that are being reported taking place more often than previously thought are dealt with to investigate if such extreme events could be forecasted in advance with artificial intelligence. Two kinds of wave time series were submitted to the present work. One is that numerically produced in a computer while linearly superposing a certain number of regular waves of different frequencies and amplitudes, in which freak-like waves appeared. The other one is that measured in a real sea collected in a 4-year project conducted by Ship Research Institute (currently National Maritime Research Institute), Japan. By courtesy of the institute, the author could access quite a large amount of digitized raw data of measured wave time series and analyze them in the present study. Quite a few freak-like waves appeared in the data. As the result of the analyses conducted in the present work using the numerically produced wave time series, it has been found that the occurrence timing of freak-like waves as well as their heights could be forecasted well with RNN. On the other hand, the results of the analyses obtained using the wave data actually measured in a real sea suggest the occurrence timing of freak-like waves could be forecasted fairly well with RNN, while the peak heights of freak-like waves predicted with RNN are noticeably lower than their actual heights. • Feasibility of forecasting a water-surface wave train with AI is investigated. • Feasibility of forecasting the occurrence of freak waves with AI is investigated. • RNN (Recurrent Neural Networks), specifically LSTM model, is used. • Freak-like waves measured in a real sea was subjected to the investigation. [ABSTRACT FROM AUTHOR]

In this paper we propose a new method for predicting rock porosity based on a combination of several artificial intelligence systems. The method focuses on one of the Iranian carbonate fields in the Persian Gulf. Because there is strong heterogeneity in carbonate formations, estimation of rock properties experiences more challenge than sandstone. For this purpose, seismic colored inversion (SCI) and a new approach of committee machine are used in order to improve porosity estimation. The study comprises three major steps. First, a series of sample-based attributes is calculated from 3D seismic volume. Acoustic impedance is an important attribute that is obtained by the SCI method in this study. Second, porosity log is predicted from seismic attributes using common intelligent computation systems including: probabilistic neural network (PNN), radial basis function network (RBFN), multi-layer feed forward network (MLFN), ε-support vector regression (ε-SVR) and adaptive neuro-fuzzy inference system (ANFIS). Finally, a power law committee machine (PLCM) is constructed based on imperial competitive algorithm (ICA) to combine the results of all previous predictions in a single solution. This technique is called PLCM-ICA in this paper. The results show that PLCM-ICA model improved the results of neural networks, support vector machine and neuro-fuzzy system. [ABSTRACT FROM AUTHOR]

Abstract: Fault location is a critical task when a severe disturbance is caused by insulation failure on a transmission line. In order to avoid further economical and social costs because of load interruptions, the fault diagnosis has to be concluded as soon as possible. Intelligent systems have been successful in dealing with fault diagnosis problems. This paper proposes the application of complex-domain neural networks for mapping the relationship between electrical signals and fault locations on transmission lines. Complex-domain neural networks allow voltage/current representation without arbitrarily decoupling amplitude and phase. Furthermore, several voltage and current representation schemes, based on electromagnetic transient and steady-state information, are analyzed in this paper. For comparison purpose, these input representations are also tested with real-domain neural networks. The tests consider realistic operating/fault conditions and assume that fault classification has already been handled. [Copyright &y& Elsevier]

Abstract: ICT-based decision aids are currently making waves in the modern business world simultaneously with increased pressure on auditors to play a more effective role in the governance and control of corporate entities. This paper aims to review the main research efforts and current debates on auditors’ use of artificial intelligent systems, with a view to predicting future directions of research and software development in the area. The paper maps the development process of artificial intelligent systems in auditing in the light of their identified benefits and drawbacks. It also reviews previous research efforts on the use of expert systems and neural networks in auditing and the implications thereof. The synthesis of these previous studies revealed certain research vacuum which future studies in the area could fill. Such areas include matching the benefits of adopting these intelligent agents with their costs, assessing the impact of artificial intelligence on internal control systems’ design and monitoring as well as audit committees’ effectiveness, and implications of using such systems for small and medium audit firms’ operations and survival, audit education, public sector organisations’ audit, auditor independence and audit expectations-performance gap. [Copyright &y& Elsevier]

Artificial Intelligence (AI) algorithms are becoming beneficial as substitute methods to conventional approaches or as components of incorporated systems. They have been utilized for solving complex applied problems in different fields and are becoming more and more popular at present. AI approaches can learn from patterns; are fault tolerant in the sense that they are capable to handle noisy data; are capable to manage non-linear problems; and once learned can carry out generalization and estimation at great speed. In this survey, for the first time, a comprehensive review of the AI algorithms developed to investigate different issues related to nanofluids is conducted. The contributions presented in this paper reveal the high potential of AI methods as tools for the prediction and optimization in nanofluids. In addition, challenges and directions for future research in the area of employing AI techniques in nanofluids are introduced and discussed. Unlabelled Image [ABSTRACT FROM AUTHOR]

In this paper we introduce the current thrust of development in machine learning and artificial intelligence, fueled by advances in statistical learning theory over the last 20 years and commercial successes by leading big data companies. Then we discuss the characteristics of process manufacturing systems and briefly review the data analytics research and development in the last three decades. We give three attributes for process data analytics to make machine learning techniques applicable in the process industries. Next we provide a perspective on the currently active topics in machine learning that could be opportunities for process data analytics research and development. Finally we address the importance of a data analytics culture. Issues discussed range from technology development to workforce education and from government initiatives to curriculum enhancement. [ABSTRACT FROM AUTHOR]

Abstract Functional networks (FNs) are extensions of neural networks (NNs). Unlike NNs, FNs considers general functional models instead of sigmoid-like models. Additionally, in FNs, there are no weights associated with the links that connect neurons. In this paper, we review the research progress and applications of FNs models in recent years. First, we introduce FNs architecture, three typical functional models and the learning process, and we explain the differences between NNs and FNs. Second, we discuss recent applications of FNs that have been introduced in many fields, such as time series prediction, differential and functional equations, pattern classification, detection and prediction, approximation computation, complex system modeling, computer-aided design (CAD), and linear and nonlinear regression. Finally, we present some remarks on future research directions for FNs. [ABSTRACT FROM AUTHOR]

Due to industrialization and the growth of transportation systems, the number of vehicles continues to increase which causes a significant traffic jam problem especially in big cities. Consequently, the prediction of traffic flows is a key component to an optimal traffic management. As a solution to this issue, the present paper aims at applying the artificial intelligence of neural networks, which offers an interesting approach to modelling in complex, and nonlinear situations. Our resolution method is based on the design of a neural network to predict daily traffic flow. Then, the forecasted traffic flow is compared with a real dataset recorded on a road section and provided by a recognized infrastructure manager in Morocco. Indeed, neural networks have the ability to learn from the past and predict the future. In this study, various neural networks structures are exanimated and simulation results show that the best forecasts are obtained with the use of Multi-Layer Perceptron architecture that has a good generalization capacity with a total Mean Square Error of 0.00927 in the train set and 0.01321 in the test set. [ABSTRACT FROM AUTHOR]

Abstract: Providing indoor thermal comfort and saving energy are two main goals of indoor environmental control system. An intelligent comfort control system by combining the intelligent control and minimum power control strategies for the indoor environment is presented in this paper. In the system, for realizing the comfort control, the predicted mean vote (PMV) is designed as the control goal, and with chastening formulas of PMV, it is controlled to optimize for improving indoor comfort lever by considering six comfort related variables. On the other hand, a RBF neural network based on genetic algorithm is designed to calculate PMV for better performance and overcoming the nonlinear feature of the PMV calculation better. The formulas given in the paper are presented for calculating the expected output values basing on the input samples, and the RBF network model is trained depending on input samples and the expected output values. The simulation result is proved that the design of the intelligent calculation method is valid. Moreover, this method has a lot of advancements such as high precision, fast dynamic response and good system performance are reached, it can be used in practice with requested calculating error. [Copyright &y& Elsevier]

Abstract: Stock price predictions have been a field of study from several points of view including, among others, artificial intelligence and expert systems. For short-term predictions, the technical indicator relative strength indicator (RSI) has been published in many papers and used worldwide. CAST is presented in this paper. CAST can be seen as a set of solutions for calculating the RSI using artificial intelligence techniques. The improvement is based on the use of feedforward neural networks to calculate the RSI in a more accurate way, which we call the iRSI. This new tool will be used in two scenarios. In the first, it will predict a market – in our case, the Spanish IBEX 35 stock market. In the second, it will predict single-company values pertaining to the IBEX 35. The results are very encouraging and reveal that the CAST can predict the given market as a whole along with individual stock pertaining to the IBEX 35 index. [Copyright &y& Elsevier]

Abstract: This paper reports the results of an advanced algorithm for the optimization of electron beam parameters in Free Electron Laser (FEL) Linacs. In the novel approach presented in this paper, the system uses state of the art developments in video games to mimic an operator''s decisions to perform an optimization task when no prior knowledge, other than constraints on the actuators is available. The system was tested for the simultaneous optimization of the energy spread and the transmission of the Australian Synchrotron Linac. The proposed system successfully increased the transmission of the machine from 90% to 97% and decreased the energy spread of the beam from 1.04% to 0.91%. Results of a control experiment performed at the new FERMI@Elettra FEL is also reported, suggesting the adaptability of the scheme for beam-based control. [Copyright &y& Elsevier]

Highlights • We survey recent works in visual analytics for deep learning engineering. • We propose a task-and-technique taxonomy to classify such approaches. • Our contribution surveys 40 papers covering multiple tasks and architectures. • We aim to offer a guide to both visual analytics and machine learning experts. • Finally, we outline challenges and topics for future research in the field. Graphical abstract Abstract Although deep neural networks have achieved state-of-the-art performance in several artificial intelligence applications in the past decade, they are still hard to understand. In particular, the features learned by deep networks when determining whether a given input belongs to a specific class are only implicitly described concerning a considerable number of internal model parameters. This makes it harder to construct interpretable hypotheses of what the network is learning and how it is learning—both of which are essential when designing and improving a deep model to tackle a particular learning task. This challenge can be addressed by the use of visualization tools that allow machine learning experts to explore which components of a network are learning useful features for a pattern recognition task, and also to identify characteristics of the network that can be changed to improve its performance. We present a review of modern approaches aiming to use visual analytics and information visualization techniques to understand, interpret, and fine-tune deep learning models. For this, we propose a taxonomy of such approaches based on whether they provide tools for visualizing a network's architecture, to facilitate the interpretation and analysis of the training process, or to allow for feature understanding. Next, we detail how these approaches tackle the tasks above for three common deep architectures: deep feedforward networks, convolutional neural networks, and recurrent neural networks. Additionally, we discuss the challenges faced by each network architecture and outline promising topics for future research in visualization techniques for deep learning models. [ABSTRACT FROM AUTHOR]

Abstract Using tools from artificial intelligence, mainly artificial neural networks, this paper presents an walking-engine for humanoid robots. This engine uses dynamic neural networks with feedback for gait generation, a modified Zhang neural network for a singularity-robust inverse kinematics solver, and feedforward neural networks for neuro-adaptive control. The Atlas humanoid robot model in simulation is used to test and verify the capabilities of the neuro-dynamic walking engine. Results show that the engine is capable of generating conventional ZMP stable walking gaits and executing them using the Atlas robot in simulation. Highlights • Develop, implement, and test the feasibility of a neuro-dynamic walking-engine. • Presents abstraction of a neural-process based walking engine. • Dynamic-model-free humanoid that is robust and can adapt to dynamic uncertainties. • Develops a novel feedback Dynamic Neural Network for gait generation. • Modified Zhang Neural Network for singularity robust Inverse Kinematics. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents a fast and reliable approach for the localization of an iris image using neural networks by comparing some of the existing localization methods. Normally any biometric recognition system is used for identification of an individual and verified with the available database to check whether the person is authorized or not. Nowadays, iris recognition systems are considered the best authentication method compared to other biometric systems due to the unique characteristic feature of the human iris. In iris recognition systems, the important and difficult step is to locate or segment the iris from the input eye image which is responsible for success rate of the iris recognition. Hence this paper suggests a efficient approach to locate the iris using neural networks in order to improve the efficiency of recognition systems. Further, the paper compares the existing iris localization methods such as Daugman algorithm, Hough transform and Canny edge detector algorithm. The best localization algorithm is chosen for training a network and simulating for efficient and fast segmentation of irises with good success rate. [Copyright &y& Elsevier]

Abstract: This paper presents a fully automatic system intended to detect leaks of dielectric fluid in underground high-pressure, fluid-filled (HPFF) cables. The system combines a number of artificial intelligence (AI) and data processing techniques to achieve high detection capabilities for various rates of leaks, including leaks as small as 15 l per hour. The system achieves this level of precision mainly thanks to a novel auto-tuning procedure, enabling learning of the Bayesian network – the decision-making component of the system – using simulated leaks of various rates. Significant new developments extending the capabilities of the original leak detection system described in and form the basis of this paper. Tests conducted on the real-life HPFF cable system in New York City are also discussed. [ABSTRACT FROM AUTHOR]

Abstract: The discrete delayed Hopfield neural networks is an extension of the discrete Hopfield neural networks. In this paper, the convergence of discrete delayed Hopfield neural networks is mainly studied, and some results on the convergence are obtained by using Lyapunov function. Several new sufficient conditions for the delayed networks converging towards a limit cycle with period at most 2 are proved in parallel updating mode. Also, some conditions for the delayed networks converging towards a limit cycle with 2-period are investigated in parallel updating mode. All results established in this paper extend the previous results on the convergence of both the discrete Hopfield neural networks, and the discrete delayed Hopfield neural networks in parallel updating mode. [Copyright &y& Elsevier]

Abstract: This paper analyses the strengths and weaknesses of self-organising approaches, such as evolutionary robotics, and direct design approaches, such as behaviour-based controllers, for the production of autonomous robots'' controllers, and shows how the two approaches can be usefully combined. In particular, the paper proposes a method for encoding evolved neural-network based behaviours into motor schema-based controllers and then shows how these controllers can be modified and combined to produce robots capable of solving new tasks. The method has been validated in the context of a collective robotics scenario in which a group of physically assembled simulated autonomous robots are requested to produce different forms of coordinated behaviours (e.g., coordinated motion, walled-arena exiting, and light pursuing). [Copyright &y& Elsevier]

Abstract: This paper discusses an experimental study of the Japanese candlestick method as used in hybrid stock market forecasting models. Two models are presented in this paper. Model 1 is a committee machine with simple generalized regression neural networks (GRNN) experts. This model also has a simple gating network. Model 2 has a similar committee machine along with a hybrid type gating network that contains fuzzy logic. Model 1 was developed to introduce the candlestick method and examine whether using the candlestick method improves performance. Model 2 is developed to determine whether the application of fuzzy logic could improve the former model. This model uses standard IF-THEN rules based fuzzy logic. In the experiment, a few simple Japanese candlestick patterns are integrated into the models. Both models use the same simple candlestick patterns to provide a basis for comparison. The Japanese candlestick method is implemented in the gating network. Model 1 uses features of candlestick patterns in the gating network. Model 2 uses candlestick patterns for recognizing the strength of market conditions. To investigate the performance of these models, the daily stock quotes of Hewlett-Packard, Bank of America, Ford, DuPont, and Yahoo are used as input data sets. The performance of the models was satisfactory based on the mean squared error. [Copyright &y& Elsevier]