Search

Artificial Intelligence (AI) has emerged as a complementary technology in supply chain research. However, the majority of AI approaches explored in this context afford little to no explainability, which is a significant barrier to a broader adoption of AI in supply chains. In recent years, the need for explainability has been a strong impetus for research in hybrid AI methodologies that combine neural architectures with logic-based reasoning, which are collectively referred to as Neurosymbolic AI. The aim of this paper is to provide a comprehensive overview of supply chain management literature that employs approaches within the neurosymbolic AI spectrum. To that end, a systematic review is conducted, followed by bibliometric, descriptive and thematic analyses on the identified studies. Our findings indicate that researchers have primarily focused on the limited subset of neurofuzzy approaches, while some supply chain applications, such as performance evaluation and sustainability, and sectors such as pharmaceutical and construction have received less attention. To help address these gaps, we propose five pillars of neurosymbolic AI research for supply chains and provide four use cases of applying unexplored neurosymbolic AI approaches to address typical problems in supply chain management, including a discussion of prerequisites for adopting such technologies. We envision that the findings and contributions of this survey will help encourage further research in neurosymbolic AI for supply chains and increase adoption of such technologies within supply chain practice. [ABSTRACT FROM AUTHOR]

Total Electron Content (TEC) forecasting using machine learning has been extensively preferred in characterizing the spatio-temporal variability of the ionosphere, to support space communication and navigation applications. Although, a variety of machine learning methods have been evolved, still, there is a greater persistence in the prediction of ionospheric TEC due to adverse space weather impacts and the complex behavior of ionosphere. And hence, more sophisticated short-term ionospheric TEC forecasting models should be developed for better prediction accuracy. The present study investigated and evaluated the performance of three models – Light Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting (XGBoost), and Gradient Boost Regression (GBR) algorithms using the data set for 25 years (1997–2021) GNSS Earth Observation Network System (GEONET) Global mean TEC time-series data. For better prediction accuracy, the data of geomagnetic storm indices Dst and Mg-II index were also utilized in training the three models. One year of data for both high (2015) and low (2020) solar activity were used to test the three models. The prediction plots are used to visualize the level of prediction error for the three algorithms. The statistical results illustrate that the LightGBM model performed better in predicting TEC with an RMSE value of 2.25 TECU (2015) and 0.52 TECU (2020) for high and low solar activity years respectively for global mean TEC data and with an RMSE of 2.34 TECU at Japan grid point location of (34.95°N and 134.05°E). In turn the XGBoost and GBR models provide an absolute TEC forecasting with an RMSE of 2.76 and 2.59 TECU (2015) and 0.60 and 0.55 TECU (2020) correspondingly for global data and as 2.39 and 2.93 TECU for grid point location. [ABSTRACT FROM AUTHOR]

In recent years, cyber security has become a increasingly prominent, with the core issue being the design of secure communication. In this paper, a secure communication framework based on secure interlayer exponential synchronous of duplex artificial neural networks (DANNs) under cyber attacks is developed. To resist random uncertainties and cyber attacks which is conceled in transmission of the signals, based on the Lyapunov's second method, an intelligent adaptive controller is designed. Based upon Lyapunov stable theory, the criteria for the interlayer synchronous are got. Finally, a secure communication framework based on the secure synchronous policy is derived.Through simulation experiments, exponential synchronous for DANNs and signal security communication have been achieved. Thus, the effectiveness of the theoretical framework is validated. The researches provide references in the design of secure communication schemes in engineering. [ABSTRACT FROM AUTHOR]

Key points While typically diagnosed with biopsy, ECRS may be predicted preoperatively with the use of AI. Various AI models have been used, with pooled sensitivity of 0.857 and specificity of 0.850. We found no statistically significant difference between the accuracy of various AI models. While typically diagnosed with biopsy, ECRS may be predicted preoperatively with the use of AI.Various AI models have been used, with pooled sensitivity of 0.857 and specificity of 0.850.We found no statistically significant difference between the accuracy of various AI models. [ABSTRACT FROM AUTHOR]

Generating high‐fidelity garment animations through traditional workflows, from modeling to rendering, is both tedious and expensive. These workflows often require repetitive steps in response to updates in character motion, rendering viewpoint changes, or appearance edits. Although recent neural rendering offers an efficient solution for computationally intensive processes, it struggles with rendering complex garment animations containing fine wrinkle details and realistic garment‐and‐body occlusions, while maintaining structural consistency across frames and dense view rendering. In this paper, we propose a novel approach to directly synthesize garment animations from body motion sequences without the need for an explicit garment proxy. Our approach infers garment dynamic features from body motion, providing a preliminary overview of garment structure. Simultaneously, we capture detailed features from synthesized reference images of the garment's front and back, generated by a pre‐trained image model. These features are then used to construct a neural radiance field that renders the garment animation video. Additionally, our technique enables garment recoloring by decomposing its visual elements. We demonstrate the generalizability of our method across unseen body motions and camera views, ensuring detailed structural consistency. Furthermore, we showcase its applicability to color editing on both real and synthetic garment data. Compared to existing neural rendering techniques, our method exhibits qualitative and quantitative improvements in garment dynamics and wrinkle detail modeling. Code is available at https://github.com/wrk226/GarmentAnimationNeRF. [ABSTRACT FROM AUTHOR]

A pivotal aspect in the design of neural networks lies in selecting activation functions, crucial for introducing nonlinear structures that capture intricate input–output patterns. While the effectiveness of adaptive or trainable activation functions has been studied in domains with ample data, like image classification problems, significant gaps persist in understanding their influence on classification accuracy and predictive uncertainty in settings characterized by limited data availability. This research aims to address these gaps by investigating the use of two types of adaptive activation functions. These functions incorporate shared and individual trainable parameters per hidden layer and are examined in three testbeds derived from additive manufacturing problems containing fewer than 100 training instances. Our investigation reveals that adaptive activation functions, such as Exponential Linear Unit (ELU) and Softplus, with individual trainable parameters, result in accurate and confident prediction models that outperform fixed-shape activation functions and the less flexible method of using identical trainable activation functions in a hidden layer. Therefore, this work presents an elegant way of facilitating the design of adaptive neural networks in scientific and engineering problems. [ABSTRACT FROM AUTHOR]

Background: Advances in medical imaging have made it possible to classify ankle fractures using Artificial Intelligence (AI). Recent studies have demonstrated good internal validity for machine learning algorithms using the AO/OTA 2018 classification. This study aimed to externally validate one such model for ankle fracture classification and ways to improve external validity. Methods: In this retrospective observation study, we trained a deep-learning neural network (7,500 ankle studies) to classify traumatic malleolar fractures according to the AO/OTA classification. Our internal validation dataset (IVD) contained 409 studies collected from Danderyd Hospital in Stockholm, Sweden, between 2002 and 2016. The external validation dataset (EVD) contained 399 studies collected from Flinders Medical Centre, Adelaide, Australia, between 2016 and 2020. Our primary outcome measures were the area under the receiver operating characteristic (AUC) and the area under the precision-recall curve (AUPR) for fracture classification of AO/OTA malleolar (44) fractures. Secondary outcomes were performance on other fractures visible on ankle radiographs and inter-observer reliability of reviewers. Results: Compared to the weighted mean AUC (wAUC) 0.86 (95%CI 0.82–0.89) for fracture detection in the EVD, the network attained wAUC 0.95 (95%CI 0.94–0.97) for the IVD. The area under the precision-recall curve (AUPR) was 0.93 vs. 0.96. The wAUC for individual outcomes (type 44A-C, group 44A1-C3, and subgroup 44A1.1-C3.3) was 0.82 for the EVD and 0.93 for the IVD. The weighted mean AUPR (wAUPR) was 0.59 vs 0.63. Throughout, the performance was superior to that of a random classifier for the EVD. Conclusion: Although the two datasets had considerable differences, the model transferred well to the EVD and the alternative clinical scenario it represents. The direct clinical implications of this study are that algorithms developed elsewhere need local validation and that discrepancies can be rectified using targeted training. In a wider sense, we believe this opens up possibilities for building advanced treatment recommendations based on exact fracture types that are more objective than current clinical decisions, often influenced by who is present during rounds. [ABSTRACT FROM AUTHOR]

This study addresses the inherent challenges associated with the limited flexibility of power systems, specifically emphasizing uncertainties in solar power due to dynamic regional and seasonal fluctuations in photovoltaic (PV) potential. The research introduces a novel supervised machine learning model that focuses on regression methods specifically tailored for advanced microgrid energy management within a 100% PV microgrid, i.e. a microgrid system that is powered entirely by solar energy, with no reliance on other energy sources such as fossil fuels or grid electricity. In this context, "PV" specifically denotes photovoltaic solar panels that convert sunlight into electricity. A distinctive feature of the model is its exclusive reliance on current solar radiation as an input parameter to minimize prediction errors, justified by the unique advantages of supervised learning. The performance of four well-established supervised machine learning models—Neural Networks (NN), Gaussian Process Regression (GPR), Support Vector Machines (SVM), and Linear Regression (LR)—known for effectively addressing short-term uncertainty in solar radiation, is thoroughly evaluated. Results underscore the superiority of the NN approach in accurately predicting solar irradiance across diverse geographical sites, including Cairo, Egypt; Riyadh, Saudi Arabia; Yuseong-gu, Daejeon, South Korea; and Berlin, Germany. The comprehensive analysis covers both Global Horizontal Irradiance (GHI) and Direct Normal Irradiance (DNI), demonstrating the model's efficacy in various solar environments. Additionally, the study emphasizes the practical implementation of the model within an Energy Management System (EMS) using Hybrid Optimization of Multiple Electric Renewables (HOMER) software, showcasing high accuracy in microgrid energy management. This validation attests to the economic efficiency and reliability of the proposed model. The calculated range of error, as the median error for cost analysis, varies from 2 to 6%, affirming the high accuracy of the proposed model. Highlights: Introduction of a novel approach: This study presents a pioneering methodology focusing exclusively on supervised machine learning techniques for short-term uncertainty management within 100% PV microgrids, aimed at optimizing energy management efficiency. Comprehensive comparative analysis: Through meticulous comparative investigation, various regression techniques for predicting solar radiation within PV microgrids are scrutinized, providing valuable insights into their efficacy across diverse environmental conditions. Validation across diverse locations: The validation of findings across four distinct geographical locations, encompassing both PV and concentrated PV (CPV) systems, substantially enhances the generalizability of the results, advancing the understanding of solar radiation prediction dynamics for renewable energy integration strategies. [ABSTRACT FROM AUTHOR]

The analysis of seismic events recorded by NASA's InSight seismometer remains challenging, given their commonly low magnitudes and large epicentral distances, and concurrently, strongly varying background noise. These factors collectively result in low signal-to-noise ratios (SNR) across most event recordings. We use a deep learning denoising approach to mitigate the noise contamination, aiming to enhance the data analysis and the seismic event catalogue. Our systematic tests demonstrate that denoising performs comparable to fine-tuned bandpass filtering at high SNRs, but clearly outperforms it at low SNRs with respect to accurate waveform and amplitude retrieval, as well as onset picking. We review the denoised waveform data of all 98 low-frequency events in the Marsquake Service catalogue version 14, and improve their location when possible through the identification of phase picks and backazimuths, while ensuring consistency with the raw data. We demonstrate that several event waveforms can be explained by marsquake doublets—two similarly strong quakes in spatio-temporal proximity that result in overlapping waveforms at InSight—and we locate them in Cerberus Fossae (CF). Additionally, we identify and investigate aftershocks and an event sequence consisting of numerous relatively high magnitude marsquakes occurring within hours at epicentral distances beyond CF. As a result of this review and interpretation, we extend the catalogue in event numbers (⁠|$+$| 8 per cent), in events with epicentral distances and magnitudes (⁠|$+$| 50 per cent), and events with backazimuths and a resulting full locations (⁠|$+$| 46 per cent), leading to a more comprehensive description of Martian seismicity. [ABSTRACT FROM AUTHOR]

The conventional transient electromagnetic inversion method has a low calculation speed and precision and is susceptible to falling into local minima, which does not meet the fine detection requirements of urban underground space. In this study, we proposed a novel inversion method based on convolutional bidirectional long short-term memory neural networks for shallow subsurface transient electromagnetic inversion. This network structure possessed strong spatial feature extraction capabilities and a proficient understanding of sequential data, thereby addressing the issues of slow conventional inversion computations and inadequate inversion accuracy. Utilizing the apparent resistivity from a three-layer model as the sample input and the real model as the target, the network was trained using batch normalization and dropout techniques to accelerate the convergence rate. The resulting model achieved real-time inversion speeds and high accuracy, with robust generalization capabilities and adaptability to new data. To assess the inversion performance, we used a novel 1-D inversion error calculation index, the correlation area loss error, for a more accurate measurement. Numerical simulation experiments showed that the proposed method required only 2.121 s to invert data from 100 observation points. The inversion efficiency was significantly superior to the conventional methods, maintaining excellent accuracy while effectively discerning subsurface electrical stratification in geophysics. Applying convolutional bidirectional long short-term memory neural networks to multidimensional and field data yielded results superior to those of conventional inversion, demonstrating the promising applicability and generalization of this approach. This study offers an efficient solution for shallow subsurface transient electromagnetic exploration and holds potential for application in other areas. [ABSTRACT FROM AUTHOR]

Radiation parameterization schemes are crucial components of weather and climate models, but they are also known to be computationally intensive. In recent decades, researchers have attempted to emulate these schemes using neural networks, with more attention to convolutional neural networks. However, in this paper, we explore the potential of recurrent neural networks (RNNs) for predicting solar heating rates. Our architecture was trained and tested using long-term hindcast data from the Pechora Sea region, with the conventional RRTMG scheme serving as a shortwave parameterization. Our findings show that the RNN offers rapid learning, fast inference, and excellent data fitting. We also present preliminary results demonstrating the use of RNNs for operational weather forecasting, which achieved a significant speedup in parameterization and reduced the overall forecast time by 40 %. [ABSTRACT FROM AUTHOR]

The motive of current work is to design a novel radial basis Bayesian regularization neural network (RB-BRNN) for solving the nonlinear fractional economic and environmental system (FEES). A radial basis activation function in the hidden layers is applied by taking 20 numbers of neurons. The mathematical FEES is presented in three classes, named as cost of control accomplishment, manufacturing elements competence and technical exclusion’s diagnostics cost. A reference dataset is obtained using the Adams numerical results to reduce the mean square error (MSE) by taking the data for training 70%, while 15% is used for both testing and validation. The negligible absolute error values and comparison of the solutions develop the worth of computing RB-BRNN in order to solve the nonlinear dynamics of the FEES. Error diagrams, regression values, and the MSE performances are implemented to assess the precision of the designed solver. [ABSTRACT FROM AUTHOR]

Uncrewed Aircraft Systems (UAS) have been recognized as powerful tools for providing monitoring data supporting mitigating and adapting to various abiotic and biotic forest disturbances. However, their suitability and commercial viability have been significantly limited by the necessity for operation primarily within visual line of sight (VLOS). Efforts are currently ongoing to establish frameworks for beyond visual line of sight (BVLOS) UAS operation. This study represents the first successful application of hydrogen-powered BVLOS airship UAS technology for monitoring forest disturbances. The test area covered 1.3 km2 of conserved forest area in North Karelia, Finland, seriously affected by bark beetles and other disturbance agents. Monitoring flights were conducted in spring, summer, and autumn in BVLOS setting operated from a command centre situated approximately 75 km distance from the test area. Deep learning models were applied to study the tree health in the disturbance area, with the special focus on examining the scalability of machine learning models. Multispectral imagery outperformed RGB imagery in classifying trees as healthy, infested, and dead spruce trees, as well as non-spruce tree classes. With multispectral imagery, the F1-scores were 0.88–0.94, 0.80, 0.92–0.96, and 0.75–0.82 for healthy, infested, and dead spruce trees, and non-spruces, respectively. Transfer learning and fine-tuning, which adapted previously trained models to new study areas, improved results compared to training models from scratch and thus contributed to the development of scalable UAS remote sensing methods in mitigating forest disturbances on a larger scale. [ABSTRACT FROM AUTHOR]

The safety monitoring for nonlinear dynamical systems with embedded neural network components is addressed in this paper. The interval-observer-based safety monitor is developed consisting of two auxiliary neural networks derived from the neural network components of the dynamical system. Due to the presence of nonlinear activation functions in neural networks, we use quadratic constraints on the global sector to abstract the nonlinear activation functions in neural networks. By combining a quadratic constraint approach for the activation function with Lyapunov theory, the interval observer design problem is transformed into a series of quadratic and linear programming feasibility problems to make the interval observer operate with the ability to correctly estimate the system state with estimation errors within acceptable limits. The applicability of the proposed method is verified by simulation of the lateral vehicle control system. [ABSTRACT FROM AUTHOR]

In this paper, the spacecraft formation control problem is investigated under the undirected tree communication topology subjected to connectivity preserving and collision avoidance constraint. To implement the connectivity preservation and collision prevention of initially connected spacecrafts without employing any potential functions, a novel appointed-time performance function is developed to impose restrictions on the relative orbit distance of the adjacent spacecrafts. Moreover, this function can also characterise the upper boundary of the convergence time and steady-state error explicitly. Then a reinforcement-learning-based relative orbit-attitude control scheme is proposed to propel the formation error to the vicinity of the origin. An actor-critic neural network architecture is utilised to online compensate the system uncertainties, in which the critic neural network is introduced to improve the approximate performance of the actor neural network for the unknown uncertainties by evaluating the consensus performance of the formation system. In addition, an auxiliary system is devised to cope with the input saturation constraint. By using the Lyapunov stability theory, the stability and convergence of the closed-loop system are analysed. Finally numerical simulations are conducted to illustrate the feasibility and effectiveness of the proposed formation control scheme. [ABSTRACT FROM AUTHOR]

This article discusses the evaluation of tool wear during the turning of Inconel 718 using self-propelled rotary tools (SPRTs) and conventional round tools (CRTs). The study aims to understand the tool wear progression and develop mathematical models to analyze the effect of cutting parameters and machining time on tool wear. The researchers conducted experiments with different cutting conditions and measured tool wear using scanning electron microscopy and digital microscopy. The results show that SPRTs have longer tool life and more uniform wear compared to CRTs. The article provides valuable information for the metalworking industry to make decisions about tool replacement and cutting conditions. [Extracted from the article]

The explosive growth in supercomputers capacity has changed simulation paradigms. Simulations have shifted from a few lengthy ones to an ensemble of multiple simulations with varying initial conditions or input parameters. Thus, an ensemble consists of large volumes of multi-dimensional data that could go beyond the exascale boundaries. However, the disparity in growth rates between storage capabilities and computing resources results in I/O bottlenecks. This makes it impractical to utilize conventional post-processing and visualization tools for analyzing such massive simulation ensembles. In situ visualization approaches alleviate I/O constraints by saving predetermined visualizations in image databases during simulation. Nevertheless, the unavailability of output raw data restricts the flexibility of post hoc exploration of in situ approaches. Much research has been conducted to mitigate this limitation, but it falls short when it comes to simultaneously exploring and analyzing parameter and ensemble spaces. In this paper, we propose an expert-in-the-loop visual exploration analytic approach. The proposed approach leverages: feature extraction, deep learning, and human expert–AI collaboration techniques to explore and analyze image-based ensembles. Our approach utilizes local features and deep learning techniques to learn the image features of ensemble members. The extracted features are then combined with simulation input parameters and fed to the visualization pipeline for in-depth exploration and analysis using human expert + AI interaction techniques. We show the effectiveness of our approach using several scientific simulation ensembles. [ABSTRACT FROM AUTHOR]

In this study, a hybrid artificial neural network (ANN)-Rao series (Rao_1, Rao_2, and Rao_3) algorithm model was developed to analyze water consumption in Istanbul province, Turkey. A multiple linear regression (MLR) model was developed and an ANN was also trained with back-propagation (BP) artificial bee colony (ABC) algorithms for comparison. Gross domestic product and population data were treated as independent variables. To test the accuracy of the presently developed hybrid model, its outputs were compared with those of ANN-BP, ANN-ABC, and MLR models. Error values calculated for the test set indicated that the ANN-Rao_3 algorithm outperformed the MLR, ANN-BP, and ANN-ABC reference models as well as ANN-Rao_1 and ANN-Rao_2 algorithms. Therefore, using the ANN-Rao_3 model, water consumption forecasts for Istanbul province were generated out to 2035 for low-, expected-, and high-water demand conditions. The model-generated forecasts indicate that the water requirements of Istanbul in 2035 will be between 1182.95 and 1399.54 million m3, with the upper-range estimates outpacing supplies. According to low and expected scenarios, there will be no problem in providing the water needs of Istanbul until 2035. However, according to high scenario, water needs of Istanbul will not be provided as of 2033.Therefore, water conservation policies should be enacted to ensure provision of the water needs of Istanbul province from 2033 onward. [ABSTRACT FROM AUTHOR]

We utilized a two-phase analysis using deep convolutional neural networks (DCNN) to create an automated technology that enabled us to detect and count saigas (Saiga tatarica) in satellite images with a resolution of 0.3–0.5 m/pixel (Eros-B 2012; 2013 and Beijing KA 2022 satellites). In the first phase, the satellite image is automatically divided into sections and checked for the presence or absence of clusters of objects (the "classification" phase). Then, during the second phase, only the fragments of the satellite image where at least one saiga was previously found are analyzed (the "detection" phase). The method was calibrated by training a neural network on the results of the preliminary processing of archival satellite images from 2012 and 2013, carried out manually by zoological experts. When we tested the DCNN work with a "confidence threshold" of 0.3, we identified 1284 saigas on the entire model satellite image, while a zoological expert manually identified 1412 saigas. For practical use and to assess the effectiveness of this method, we counted saigas on a 2022 image covering two adjacent specially protected natural areas (PAs) located in the Republic of Kalmykia and the Astrakhan region (Russian Federation). The results are presented with different "thresholds of confidence." [ABSTRACT FROM AUTHOR]

This research explores the incorporation of variable order (VO) fractional calculus into the hyperchaotic Lorenz system and studies various chaotic features and attractors. Initially, we propose a variable fractional order hyperchaotic Lorenz system and numerically solve it. The solutions are obtained for multiple choices of control parameters, and these results serve as reference solutions for exploring chaos with the artificial intelligence tool radial basis function network (RBFN). We rebuild phase spaces and trajectories of system states to exhibit chaotic behavior at various levels. To further assess the sensitivity of chaotic attractors, Lyapunov exponents are calculated. The efficacy of the designed computational RBFN is validated through the RMSE and extensive error analysis. The proposed research on AI capabilities aims to introduce an innovative methodology for modeling and analyzing hyperchaotic dynamical systems with variable orders. [ABSTRACT FROM AUTHOR]

In the rapidly evolving landscape of cyber threats, effective defense strategies are crucial for safeguarding sensitive information and critical systems. Deep learning methods, notably the Transformer architecture, have shown immense potential in addressing cybersecurity challenges. However, customizing, and adapting Transformer architectures for cybersecurity applications presents a challenge, demanding the utilization of effective strategies to achieve optimal performance. This study presents a comprehensive analysis of design tactics employed in tailoring Transformer architectures specifically for cybersecurity problems. Design tactics, defined as strategic solutions to architectural challenges based on well-justified design decisions, are explored in-depth within the context of cybersecurity. By examining the modifications and adaptations made to the original Transformer architecture, this study unveils the design decisions and strategies crucial for successful implementation in diverse cybersecurity domains. The findings emphasize the significance of aligning design tactics with the unique business requirements and quality factors of each specific application domain. This study contributes valuable insights into the utilization of design tactics for customizing Transformer architectures in cybersecurity, paving the way for enhanced defense strategies against the dynamic and evolving nature of cyber threats. [ABSTRACT FROM AUTHOR]

The efficacy of meta-heuristic algorithms has been demonstrated in solving unconstrained optimization problems. Inspired by the food foraging behavior of beetles, we propose herein a model-free meta-heuristic optimization algorithm, referred to as the Beetle Antennae Search-Bellwether Swarm, which incorporates the schemes of adaptive antenna fiber length and direction vector. A joint annealing-heating scheme along with the re-initialization mechanism is proposed such that each antenna fiber length could simultaneously gradually decrease or increase in different circumstances. We also propose to adapt each individual antenna fiber length based on the best smell perception direction (bellwether). The proposed schemes of adaptive antenna fiber length yield the algorithm almost insusceptible to different initial antenna fiber lengths for a relatively large range of initial antenna fiber lengths. The softmax combiner is leveraged to obtain the direction vector such that the beetle would be steered to step toward the new randomly combined direction and arrive at a candidate position. The beetle position is further updated by evaluating whether the candidate position is conducive to minimizing the objective function. The illustrative simulations demonstrate the fast initial convergence, cost-effectiveness and feasibility of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Oncolytic viral immunotherapy is gaining considerable prominence in the realm of chronic diseases treatment and rehabilitation. Oncolytic viral therapy is an intriguing therapeutic approach due to its low toxicity and dual function of immune stimulations. This work aims to design a soft computing approach using stupendous knacks of neural networks (NNs) optimized with Bayesian regularization (BR), i.e. NNs-BR, procedure. The constructed NNs-BR technique is exploited in order to determine the approximate numerical treatment of the nonlinear multi-delayed tumor virotherapy (TVT) models in terms of the dynamic interactions between the tumor cells free of viruses, tumor cells infected by viruses, viruses, and cytotoxic T-lymphocytes (CTLs). The strength of state-of-the-art numerical approach is incorporated to develop the reference dataset for the variation in the infection rate for tumor cells, virus-free tumor cell clearance rate by CTLs, CTLs clearance rate for infectious tumor cells, the natural lifecycle of infectious tumor cells, the natural lifecycle of viral cell, the natural lifecycle of CTLs cells, tumor cells free of viruses' maximum proliferation rate, production of tumor cells with an infection, CTLs simulated ratio for infectious tumor cells, CTLs simulated ratios for virus-free cells and delay in time. The dataset is randomly chosen/segmented for training-testing-validation samples to construct the NNs models optimized with backpropagated BR representing the approximate numerical solutions of the dynamic interactions in the TVT model. The performance of the designed NNs-BR technique is accessed/evaluated and outcomes are found in good agreement with the reference solutions having the range of accuracy from 10 − 9 to 10 − 1 6 . The efficacy of NNs-BR paradigm is further substantiated after rigorous analysis on regression metrics, learning curves on MSE, and error histograms for the dynamics of TVT model. [ABSTRACT FROM AUTHOR]

For switched under-actuated systems subject to bouc-wen hysteresis, an adaptive neural predefined-time control scheme is presented first time in this paper. To improve the robustness and response rate of the system under consideration, a hierarchical sliding mode control technique is introduced. Based on a predefined-time stability criterion, the stabilization time of the system can be directly preset under the designed controller. Besides, the convergence time depends only on the preset value and is independent of all other parameters. The projection algorithm is introduced to avoid the singularity problem in the controller design process. Meanwhile, it is proved that all signals of the closed-loop system are bounded after a preset time through the Lyapunov stability theory. Finally, a simulation example and a comparison example are given to further illustrate the effectiveness and superiority of developed control approach. [ABSTRACT FROM AUTHOR]

When investigating supermarket refrigeration systems, it is essential to consider the refrigeration load as it directly affects the required compressor capacity and energy consumption. Therefore, it is important to have knowledge of the refrigeration load and its main influencing factors. Direct usage of data for the refrigeration load in a whole-year energy simulation can present significant challenges. This is partly due to long lead times for data collection and the challenges posed by missing data. As a result, often there is no complete data set of the refrigeration load for the whole-year. Therefore, a mathematical model can serve as a tool to interpolate and extrapolate the refrigeration load over the course of a year. The model is not intended for transfer to other locations. This paper begins with a overview of correlations found in the literature, followed by the development of an improved model that accounts for the impact of ambient temperature and time on the refrigeration load. Additionally, a novel approach using a neural network is proposed as a second model. Both models are then applied to data from three supermarkets in Europe and the model performance is evaluated. Both models performed acceptably in representing the refrigeration load of all three supermarkets. Especially including the influence of time of day into the first model significantly improves the prediction accuracy. The neural network has the highest prediction performance and can improve the prediction performance by 40% over the improved literature model. [ABSTRACT FROM AUTHOR]

Intensity measures based on energy have proven to be robust indicators of damage for a variety of structural types. This article presents a modified ground-motion model (GMM) incorporating a deep neural network to predict elastic input energy spectra for both horizontal and vertical ground motions, considering the pulselike ground motions. The newly developed model employs six predictor variables, that is, moment magnitude MW, fault mechanism F, rupture distance Rrup, logarithmic rupture distance ln(Rrup), rupture directivity term Idir, and logarithmic shear-wave velocity ln(VS30)as inputs. A subset of records, sourced from the recently updated Next Generation Attenuation-West2 Project database constituted by 2745 ground motions from 97 earthquakes, have been employed in the development of the model. The performance of the developed model remains within the prescribed error range. In addition, the proposed model is compared against currently used GMMs. The predicted spectra obtained from the present study are in good agreement with those given by other literature, and the standard deviations of residuals have been reduced by ~20% and are more stable. Observations from these results indicate that the newly proposed model generates improved predictions. [ABSTRACT FROM AUTHOR]

Determining the importance of major antecedents of open innovation between such distinct partners as industry and universities influences the decision-making regarding resources and effort allocation to their improvement, according to the strategic objectives of the firms. For this purpose, the present paper proposes an approach for conducting their importance-performance analysis based on fuzzy set theory and neural networks. Considering a hierarchical component model that integrates the components of the major antecedents, this study advances a research framework that first involves the operationalization of the collected data as fuzzy numbers. Then, the SHapley Additive exPlanation-based method estimates the derived importance of each component in the hierarchical component model using an optimal two-layers back-propagation network. Finally, a nine quadrants division of the importance-performance analysis developed on the basis of relevance and determinance measures of the analyzed antecedent components, delineates the prioritization of their potential improvements. A case study aims to demonstrate the developed research framework, illustrating its effectiveness and flexibility in decision-making related to the improvement of such antecedents. [ABSTRACT FROM AUTHOR]

Abstract—An analysis of current trends and opportunities for the application of artificial intelligence (AI) in materials science and concrete technology, including 3D printing in construction, is presented. The key role of AI in predicting material properties, developing new materials, and quality control is highlighted. By analyzing large volumes of data collected from numerous studies, AI can suggest optimal parameters to achieve desired material properties, thereby reducing costs and increasing production efficiency. Existing rheological models, such as the Bingham–Shvedov model or the Herschel–Bulkley model, describe material behavior based on specific equations and parameters. These models can be useful in predicting concrete properties, especially when data on its component composition is available. However, these models may be limited in their predictive accuracy, particularly for nonstandard or novel materials. It has been found that machine learning and neural networks have the potential to provide accurate predictions of rheological and physicomechanical properties of concrete materials, considering multiple parameters that influence material characteristics, including chemical and mineralogical composition, as well as structural features. The combination of experimental data and AI can successfully optimize compositions and properties during production, reducing costs and research/testing time, and opening new opportunities for researchers and engineers in the field of materials science. Machine-learning algorithms such as XGBoost, LightGBM, Catboost, and NGBoost demonstrate high predictive accuracy and have become powerful tools in the design of concrete compositions and innovative technologies. The analysis of Shapley additive explanations allows us to understand which parameters of a concrete mixture have the greatest influence on its characteristics. [ABSTRACT FROM AUTHOR]

Activation functions (AFs) play a critical role in artificial neural networks, allowing for the modeling of complex, non-linear relationships in data. In this review paper, we provide an overview of the most commonly used AFs in deep learning. In this comparative study, we survey and compare the different AFs in deep learning and artificial neural networks. Our aim is to provide insights into the strengths and weaknesses of each AF and to provide guidance on the appropriate selection of AFs for different types of problems. We evaluate the most commonly used AFs, including sigmoid, tanh, rectified linear units (ReLUs) and its variants, exponential linear unit (ELU), and SoftMax. For each activation category, we discuss its properties, mathematical formulation (MF), and the benefits and drawbacks in terms of its ability to model complex, non-linear relationships in data. In conclusion, this comparative study provides a comprehensive overview of the properties and performance of different AFs, and serves as a valuable resource for researchers and practitioners in deep learning and artificial neural networks. [ABSTRACT FROM AUTHOR]

We studied the learnability of English filler-gap dependencies and the "island" constraints on them by assessing the generalizations made by autoregressive (incremental) language models that use deep learning to predict the next word given preceding context. Using factorial tests inspired by experimental psycholinguistics, we found that models acquire not only the basic contingency between fillers and gaps, but also the unboundedness and hierarchical constraints implicated in the dependency. We evaluated a model's acquisition of island constraints by demonstrating that its expectation for a filler-gap contingency is attenuated within an island environment. Our results provide empirical evidence against the argument from the poverty of the stimulus for this particular structure. [ABSTRACT FROM AUTHOR]

Given an orthographic transcription, forced alignment systems automatically determine boundaries between segments in speech, facilitating the use of large corpora. In the present paper, we introduce a neural network-based forced alignment system, the Mason-Alberta Phonetic Segmenter (MAPS). MAPS serves as a testbed for two possible improvements we pursue for forced alignment systems. The first is treating the acoustic model as a tagger, rather than a classifier, motivated by the common understanding that segments are not truly discrete and often overlap. The second is an interpolation technique to allow more precise boundaries than the typical 10 ms limit in modern systems. During testing, all system configurations we trained significantly outperformed the state-of-the-art Montreal Forced Aligner in the 10 ms boundary placement tolerance threshold. The greatest difference achieved was a 28.13 % relative performance increase. The Montreal Forced Aligner began to slightly outperform our models at around a 30 ms tolerance. We also reflect on the training process for acoustic modeling in forced alignment, highlighting how the output targets for these models do not match phoneticians' conception of similarity between phones and that reconciling this tension may require rethinking the task and output targets or how speech itself should be segmented. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) has significantly impacted various fields. Large language models (LLMs) like GPT-4, BARD, PaLM, Megatron-Turing NLG, Jurassic-1 Jumbo etc., have contributed to our understanding and application of AI in these domains, along with natural language processing (NLP) techniques. This work provides a comprehensive overview of LLMs in the context of language modeling, word embeddings, and deep learning. It examines the application of LLMs in diverse fields including text generation, vision-language models, personalized learning, biomedicine, and code generation. The paper offers a detailed introduction and background on LLMs, facilitating a clear understanding of their fundamental ideas and concepts. Key language modeling architectures are also discussed, alongside a survey of recent works employing LLM methods for various downstream tasks across different domains. Additionally, it assesses the limitations of current approaches and highlights the need for new methodologies and potential directions for significant advancements in this field. [ABSTRACT FROM AUTHOR]

Purpose: To develop neural network (NN)‐based quantitative MRI parameter estimators with minimal bias and a variance close to the Cramér–Rao bound. Theory and Methods: We generalize the mean squared error loss to control the bias and variance of the NN's estimates, which involves averaging over multiple noise realizations of the same measurements during training. Bias and variance properties of the resulting NNs are studied for two neuroimaging applications. Results: In simulations, the proposed strategy reduces the estimates' bias throughout parameter space and achieves a variance close to the Cramér–Rao bound. In vivo, we observe good concordance between parameter maps estimated with the proposed NNs and traditional estimators, such as nonlinear least‐squares fitting, while state‐of‐the‐art NNs show larger deviations. Conclusion: The proposed NNs have greatly reduced bias compared to those trained using the mean squared error and offer significantly improved computational efficiency over traditional estimators with comparable or better accuracy. [ABSTRACT FROM AUTHOR]

A signature is a biometric attribute commonly used for identity verification. It can be represented by a shape created with a classic pen, but it can also contain dynamic information. This information is acquired using a digital input device, such as a graphic tablet or a digital screen and stylus. Information about the dynamics of the signing process is stored in the form of signals that change over time, including pen velocity, pressure, and more. These dynamics are characteristic of an individual and are difficult for a human to forge. However, it is an interesting research issue whether the values of signals describing a dynamic signature can be predicted using artificial intelligence methods. Predicting the dynamics of the signals describing a signature would benefit various scientific problems, including improving the quality of reference signals by detecting anomalies, creating signature templates better suited to individuals, and more effectively detecting potential forgeries by identity verification systems. In this paper, we propose a method for predicting dynamic signature signals using an artificial neural network. The method was evaluated using samples collected in the DeepSignDB database, distributed by BiDA Lab. [ABSTRACT FROM AUTHOR]

Neural network applications, as an emerging machine learning technology, are increasingly being integrated into pharmaceutical manufacturing technologies, offering significant improvement opportunities for performance, efficiency and sustainability. This review paper utilizes a systematic methodology to establish key literature trends and themes. The state-of-the-art body of knowledge in this hot research area is analyzed in descriptive (e.g. neural network technologies studied, sustainability indicators considered, manufacturing process addressed) and thematic synthesis components. Process analysis and improvement, quality control and additive manufacturing were identified as the three focal research themes, and research lines within these themes were further studied and discussed. To guide future research, potential paths and research questions are proposed against the gaps identified. The originality of this work lies in its methodology (adoption of a systematic review approach, highly limited in the current literature), its inclusion of sustainability (as an imperative concept for manufacturing technology research) and its specific focus on neural network applications in the context of pharmaceutical manufacturing technologies (a perspective, either has been missing or addressed too widely by extant contributions). [ABSTRACT FROM AUTHOR]

Feature selection in clustering is a hard task which involves simultaneously the discovery of relevant clusters as well as relevant variables with respect to these clusters. While feature selection algorithms are often model-based through optimised model selection or strong assumptions on the data distribution, we introduce a discriminative clustering model trying to maximise a geometry-aware generalisation of the mutual information called GEMINI with a simple ℓ 1 penalty: the Sparse GEMINI. This algorithm avoids the burden of combinatorial feature subset exploration and is easily scalable to high-dimensional data and large amounts of samples while only designing a discriminative clustering model. We demonstrate the performances of Sparse GEMINI on synthetic datasets and large-scale datasets. Our results show that Sparse GEMINI is a competitive algorithm and has the ability to select relevant subsets of variables with respect to the clustering without using relevance criteria or prior hypotheses. [ABSTRACT FROM AUTHOR]

This paper examines the role of dietary peptides gluten and casein in modulating brain function in individuals with autism spectrum disorder (ASD) from a biochemical perspective. Neurotransmitter systems and neural networks are crucial for brain function, and alterations at the biochemical level can contribute to the characteristic symptoms and behaviors of ASD. The paper explores how dietary peptides influence neurotransmitter systems and neural networks, highlighting their potential as interventions to improve brain function in ASD. The evidence suggests that dietary peptides can impact neurotransmitter synthesis, release, and receptor interactions, disrupting the balance of neurotransmitter systems and affecting neural network function. The findings underscore the potential of dietary interventions in modulating brain function in ASD and call for further research to elucidate the underlying mechanisms and optimize clinical practice. Considering individual dietary sensitivities and preferences, personalized dietary approaches may be necessary for optimal outcomes. Dietary interventions' timing, duration, and integration with other evidence-based treatments are crucial considerations. Safety considerations and regular monitoring are important to ensure the implementation of dietary interventions safely and effectively. [ABSTRACT FROM AUTHOR]

Credit card fraud is a huge challenge in the financial sector, causing huge losses every year. The problem is exacerbated by increased marketing and sophisticated fraudulent activities. This study addresses the important issue of accurate real-time detection of fraudulent transactions to minimize financial losses and enhance transactional security. The main objective of this study is to develop a comprehensive fraud detection algorithm using deep learning techniques, specially designed to address the complexity and volume of modern credit card transactions. Key contributions of this research include the presentation of a new deep learning algorithm optimized for credit card fraud detection, the integration of feature engineering techniques to improve the performance of the model, and a potential scalable solution analysis in real-time Significant improvement in proven rates. The results show that the proposed deep learning-based model achieves higher accuracy and lower false positive rate, giving financial institutions a significant advantage in protecting against fraudulent activities about the character. This study highlights the power of deep learning in reforming fraud detection systems, and lays the foundation for future developments in this important area. [ABSTRACT FROM AUTHOR]

Forecasting the stock market is a significant challenge in the financial industry due to its time series' complicated, noisy, chaotic, dynamic, volatile, and non-parametric nature. Nevertheless, due to computer advancements, an intelligent model can assist investors and expert analysts mitigate the risk associated with their investments. In recent years, substantial research has been conducted on deep learning models. Many studies have investigated using these techniques to anticipate stock values by analyzing historical data and technical indications. However, since the goal is to create predictions for the financial market, validating the model using profitability indicators and model performance is crucial. This article incorporates the attention mechanism model, incorporating attention from both feature and time perspectives. Utilize artificial neural networks. This approach addresses issues in time series prediction. The issue is the varying degrees of influence that many input features have on the target sequence. To tackle this, the method utilizes a feature attention mechanism to obtain the weights of distinct input features. An enhanced feature association relationship is achieved, whereas the data before and following the sequence exhibit a significant time correlation. An attention technique is employed to address this issue, allowing for the acquisition of weights at various time intervals to enhance robustness and temporal dependence. The system is applied to the three global SMs (TESLA, S&P500, and NASDAQ) datasets, the best enhancement results are 99% in Acc, and the better results improvement to minimize error in MSE, MAPE, and RMSE are 0.004, 0.004 and 0.01 respectively. [ABSTRACT FROM AUTHOR]

In this paper, we design a neural network architecture to approximate the weakly efficient frontier of convex vector optimization problems (CVOP) satisfying Slater's condition. The proposed machine learning methodology provides both an inner and outer approximation of the weakly efficient frontier, as well as an upper bound to the error at each approximated efficient point. In numerical case studies we demonstrate that the proposed algorithm is effectively able to approximate the true weakly efficient frontier of CVOPs. This remains true even for large problems (i.e., many objectives, variables, and constraints) and thus overcoming the curse of dimensionality. [ABSTRACT FROM AUTHOR]

Device-to-Device (D2D) relayed communication helps in extending the coverage range of cellular networks. Relay devices support multi-hop D2D communication where two devices are out of the direct D2D range. However, identifying suitable fixed relays in a network is a complex problem that needs a more efficient solution. Relay-assisted communication may also fail due to the non-cooperative nature of the users (draining battery energy for supporting other devices). This paper proposes a UAV (Unmanned Aerial Vehicle)-assisted multi-hop D2D communication scheme that serves more out-of-direct-range D2D users using the dynamic location of the UAVs (drones). Dynamic location of UAVs solves the connectivity issues with many users. We aim at maximizing the achievable throughput of the D2D users for both uplink (users to UAVs) and downlink (UAVs to users) channels simultaneously. An optimization problem is formulated for maximizing throughput subject to interference, power, and bandwidth constraints. The UAV trajectories are predicted for serving the multi-hop D2D users in the system using Neural Network (NN), and thereafter, a novel resource assignment scheme, named Dual Optimal Channel Allocation (DOCA), is proposed. DOCA optimally allocates resource blocks (RBs) for both uplink and downlink channels and ensures that the overall interference caused by resource sharing between cellular and D2D users is minimal. The spectrum efficiency has been achieved by resource sharing between cellular and D2D users. An association matrix is obtained that indicates potential resource-sharing partners of D2D and cellular users. Finally, we show the performance of the proposed technique with regard to throughput improvement, buffer requirement, and churn rate of the system. [ABSTRACT FROM AUTHOR]

This survey introduces 101 new publications on applications of Clifford's geometric algebras (GAs) newly published during 2022 (until mid‐January 2023). The selection of papers is based on a comprehensive search with Dimensions.ai, followed by detailed screening and clustering. Readers will learn about the use of GA for mathematics, computation, surface representations, geometry, image, and signal processing, computing and software, quantum computing, data processing, neural networks, medical science, physics, electric engineering, control and robotics. [ABSTRACT FROM AUTHOR]

In this paper, the existence of positive periodic solutions of impulsive neutral‐type neural networks with distributed delays and two parameters is discussed by applying the cone‐theoretic fixed point theorem. Some sufficient conditions are obtained to ensure the existence of multiple and single positive periodic solutions for the considered system. Several existence results are established. Finally, we exhibit an example to illustrate the applicability of the established results. [ABSTRACT FROM AUTHOR]

In this paper, we present a variant of our previous research on multi-goal path finding problem, focusing on finding a feasible and closed path to visit a sequence of goals in an environment with obstacles. The newly proposed method, Segmentation & Regression v2 (S&Reg v2), employs multi-task learning networks to generate regions and estimates of lengths of local paths between pairwise goals. Importantly, the estimates are performed as weights for a complete graph to compute the visiting sequence. Subsequently, the path-finding process is executed following the sequence, and the predicted region works as a sampling domain to enhance the search speed. A hybrid sampler is designed by combining a uniform domain with the region domain, ensuring successful samples, even if the region is disconnected. Besides, a selection rule is introduced to balance the sampling domain during different searching stages. A proof of probabilistic completeness of the S&Reg v2 method is given. Simulations verify the superior performance of the S&Reg v2 method, demonstrating a reduction in calculation time ranging from 3.9% to 13.0%. Furthermore, a practical scenario validates the reliability of S&Reg v2, achieving a 15.0% improvement in success rate and a 9.7% reduction in calculation time. [ABSTRACT FROM AUTHOR]

Objectives. This study investigates the possibility of developing a unique model for predicting work-related injuries in Serbian underground coal mines using neural networks and fuzzy logic theory. Accidents are common due to the unique nature of underground mineral extraction involving people, machinery and limited workplaces. Methods. A universal model for predicting occupational accidents takes into account influential factors such as organizational aspects, personal and collective protective equipment, on-the-job training and leadership factors. The selected networks achieved a prediction accuracy of >90%. Results. The study successfully identifies potential risks and critical worker groups leading to injuries. The sensitivity analysis provides insights for targeted safety measures and improved organizational practices. Conclusion. This data-driven approach makes a valuable contribution to safety in the mining industry. Implementation of the predictive model can reduce injuries and machine damage, and improve worker well-being. [ABSTRACT FROM AUTHOR]

An important development in the study of face impressions was the introduction of dominance and trustworthiness as the primary and potentially orthogonal traits judged from faces. We test competing predictions of recent accounts that address evidence against the independence of these judgements. To this end we develop a version of recent ‘deep models of face impressions’ better suited for data‐efficient experimental manipulation. In Study 1 (N = 128) we build impression models using 15 times less ratings per dimension than previously assumed necessary. In Study 2 (N = 234) we show how our method can precisely manipulate dominance and trustworthiness impressions of face photographs and observe how the effects' pattern of the cues of one trait on impressions of the other differs from previous accounts. We propose an altered account that stresses how a successful execution of the two judgements' functional roles requires impressions of trustworthiness and dominance to be based on cues of both traits. Finally we show our manipulation resulted in larger effect sizes using a broader array of features than previous methods. Our approach lets researchers manipulate face stimuli for various face perception studies and investigate new dimensions with minimal data collection. [ABSTRACT FROM AUTHOR]