Search

Optimized scheduling is an important task in the scheduling of job execution on cloud virtual machines, where optimal resource usage and a shorter makespan have become important features. When scheduling jobs, balancing the workload across all available virtual machines provides optimized performance. Reinforcement learning is a better optimization algorithm due to its adaptability to dynamic environments and balancing exploration and exploitation. To perform optimized balancing of job scheduling, an Actor- Critic-based reinforcement algorithm is applied in this work. The Alibaba cloud dataset is used to analyze the algorithm's performance. Policy constraints are made for assigning the number of tasks to the scheduler. During the learning phase, the rewards turn out to be negative. After the learning phase, the rewards stabilize. The results show that the algorithm is able to produce positive reward points. A 5% reduction in the makespan of job execution demonstrates the improvement in scheduling and resource use. [ABSTRACT FROM AUTHOR]

Copyright of Remote Sensing for Natural Resources is the property of Remote Sensing for Natural Resources Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper addresses the challenges of traditional spindle assembly methods, which rely on trial-and-error approaches, hindering new product development evaluation. A stacking ensemble model is proposed to predict the assembly quality variation of machine tool spindles. The model uses data from 925 single-spindle inspections and extracts evaluation metrics from multiple domains to extract valuable information. Feature selection is performed using a correlation model to identify important features, and various lightweight supervised learning algorithms are applied to analyze the data. To further enhance the model's performance, a stacking ensemble approach is proposed, which combines algorithms. The results demonstrate that the proposed stacking ensemble model is an effective approach for predicting the assembly quality variation of machine tool spindles, using the data available. The proposed ensemble model enhances quality control processes in spindle assembly, enabling practitioners to identify key features and predict machine tool spindle assembly quality variations more accurately. [ABSTRACT FROM AUTHOR]

Efficient Learning Algorithms for the Best Capped Base-Stock Policy in Lost Sales Inventory Systems Periodic review, lost sales inventory systems with lead times are notoriously challenging to optimize. Recently, the capped base-stock policy, which places orders to bring the inventory position up to the order-up-to level subject to the order cap, has demonstrated exceptional performance. In the paper "UCB-Type Learning Algorithms with Kaplan–Meier Estimator for Lost Sales Inventory Models with Lead Times," Lyu, Zhang, and Xin propose an upper confidence bound–type learning framework. This framework, which incorporates simulations with the Kaplan–Meier estimator, works with censored demand observations. It can be applied to determine the optimal capped base-stock policy with a tight regret with respect to the planning horizon and the optimal base-stock policy with a regret that matches the best existing result. Both theoretical analysis and extensive numerical experiments demonstrate the effectiveness of the proposed learning framework. In this paper, we consider a classic periodic-review lost-sales inventory system with lead times, which is notoriously challenging to optimize with a wide range of real-world applications. We consider a joint learning and optimization problem in which the decision maker does not know the demand distribution a priori and can only use past sales information (i.e., censored demand). Departing from existing learning algorithms on this learning problem that require the convexity property of the underlying system, we develop an upper confidence bound (UCB)-type learning framework that incorporates simulations with the Kaplan–Meier estimator and demonstrate its applicability to learning not only the optimal capped base-stock policy in which convexity no longer holds, but also the optimal base-stock policy with a regret that matches the best existing result. Compared with a classic multi-armed bandit problem, our problem has unique challenges because of the nature of the inventory system, because (1) each action has long-term impacts on future costs, and (2) the system state space is exponentially large in the lead time. As such, our learning algorithms are not naive adoptions of the classic UCB algorithm; in fact, the design of the simulation steps with the Kaplan–Meier estimator and averaging steps is novel in our algorithms, and the confidence width in the UCB index is also different from the classic one. We prove the regrets of our learning algorithms are tight up to a logarithmic term in the planning horizon T. Our extensive numerical experiments suggest the proposed algorithms (almost) dominate existing learning algorithms. We also demonstrate how to select which learning algorithm to use with limited demand data. Supplemental Material: The online appendix is available at https://doi.org/10.1287/opre.2022.0273. [ABSTRACT FROM AUTHOR]

In the manufacturing industry, production efficiency is the core competitiveness of enterprises, but also an important goal of enterprise production planning management, and in recent years, the production efficiency of China's manufacturing industry does not meet the requirements, and there are still many problems to be solved. In the face of this situation, the relevant fields have explored intelligent manufacturing and proposed intelligent manufacturing technology, which mainly integrates information technology with traditional manufacturing system, can realize the important measures of intelligence, automation, digitalization and information in the manufacturing process, and can help solve the current key problems. Therefore, this paper takes ship manufacturing as the background. The real-time optimization technology of intelligent ship manufacturing based on learning algorithm is studied. This technology can effectively improve ship production efficiency and reduce production cost. [ABSTRACT FROM AUTHOR]

The long-time test of battery discharge time is the main problem for implementing the performance requirements of the standard IEC 60086. Prediction of battery discharge time can be a solution for the long-time test problem. The discharge test data conceive valuable information of the nonlinear characteristics of the battery to build battery models. An intelligent system with learning algorithm to update the capability of the model from a set of training test data is suitable to predict the battery discharge time. The well-known ANFIS (Adaptive Neuro Fuzzy Inference Systems) is a neuro-adaptive system that has adaptive input nodes and has been applied to many prediction problems through learning and validation of the network from data. In this paper, the empirical test data related to the discharge time are acquired from the laboratory tests and are used to build the persuasive ANFIS model for the discharge time prediction from inputs to 48 recombination output data by means of modified membership functions. [ABSTRACT FROM AUTHOR]

Artificial neural networks (ANNs) have led to landmark changes in many fields, but they still differ significantly from the mechanisms of real biological neural networks and face problems such as high computing costs, excessive computing power, and so on. Spiking neural networks (SNNs) provide a new approach combined with brain-like science to improve the computational energy efficiency, computational architecture, and biological credibility of current deep learning applications. In the early stage of development, its poor performance hindered the application of SNNs in real-world scenarios. In recent years, SNNs have made great progress in computational performance and practicability compared with the earlier research results, and are continuously producing significant results. Although there are already many pieces of literature on SNNs, there is still a lack of comprehensive review on SNNs from the perspective of improving performance and practicality as well as incorporating the latest research results. Starting from this issue, this paper elaborates on SNNs along the complete usage process of SNNs including network construction, data processing, model training, development, and deployment, aiming to provide more comprehensive and practical guidance to promote the development of SNNs. Therefore, the connotation and development status of SNN computing is reviewed systematically and comprehensively from four aspects: composition structure, data set, learning algorithm, software/hardware development platform. Then the development characteristics of SNNs in intelligent computing are summarized, the current challenges of SNNs are discussed and the future development directions are also prospected. Our research shows that in the fields of machine learning and intelligent computing, SNNs have comparable network scale and performance to ANNs and the ability to challenge large datasets and a variety of tasks. The advantages of SNNs over ANNs in terms of energy efficiency and spatial-temporal data processing have been more fully exploited. And the development of programming and deployment tools has lowered the threshold for the use of SNNs. SNNs show a broad development prospect for brain-like computing. [ABSTRACT FROM AUTHOR]

Predicting time series, especially those originating from chaotic and nonlinear dynamic systems, is a critical research area with broad applications across various fields. Neural networks and fuzzy systems have emerged as leading methods for forecasting chaotic time series. This study introduces an improved adaptive neural-fuzzy inference system (ANFIS) specifically tailored for forecasting chaotic time series. Unlike traditional ANFIS models, which are primarily designed for static problems, this enhanced version incorporates self-feedback relationships from previous outputs to capture the time dependencies inherent in dynamic systems. Additionally, a hybrid approach combining the Imperialist Competitive Optimization Algorithm (ICA) and Least Squares Estimation (LSE) is employed to train the neural-fuzzy system and update its parameters. This method circumvents challenges associated with training gradient-based algorithms. The proposed technique is applied to predict and model multiple nonlinear and chaotic time series from realworld scenarios. Comparative analyses with recent works demonstrate the superior performance of the proposed method, particularly in terms of the prediction total error criterion for time series modeling and forecasting. These results highlight the effectiveness of incorporating self-feedback relationships and utilizing the CCA-LSE hybrid approach in enhancing the predictive capabilities of adaptive neural-fuzzy inference systems for chaotic time series. [ABSTRACT FROM AUTHOR]

Spiking neural network (SNN) is a new generation of artificial neural network. It is more biologically plausible and has been widely concerned by scholars owing to its unique information coding schemes, rich spatiotemporal dynamics, and event-driven operating mode with low power. In recent years, SNN has been explored and applied in many fields such as medical health, industrial detection, and intelligent driving. First, the basic elements and learning algorithms of SNN are introduced, including classical spiking neuron models, spike-timing dependent plasticity (STDP), and common information coding methods. The advantages and disadvantages of the learning algorithms are also analyzed. Then, the mainstream software simulators and neuromorphic hardware of SNN are summarized. Subsequently, the research progress and application scenarios of SNN in terms of computer vision, natural language processing, and reasoning decision are introduced. Particularly, SNN has shown strong potentials in tasks such as object detection, action recognition, semantic cognition, and speech recognition, significantly improving computational performance. Future research and application of SNN should focus on strengthening the research on key core technologies, promoting the application of technological achievements, and continuously optimizing the industrial ecology, thus to catch up with the advanced international level. Moreover, continuous research and breakthroughs of brain-inspired systems and control theories will promote the establishment of large-scale SNN models and are expected to broaden the application prospect of artificial intelligence.

Abstract This ongoing work is vehemently dedicated to the investigation of a class of ordinary linear Volterra type integro-differential equations with fractional order in numerical mode. By replacing the unknown function by an appropriate multilayered feed-forward type neural structure, the fractional problem of such initial value is changed into a course of non-linear minimization equations, to some extent. Put differently, interest was sparked in structuring an optimized iterative first-order algorithm to estimate solutions for the origin fractional problem. On top of that, some computer simulation models exemplify the preciseness and well-functioning of the indicated iterative technique. The outstanding accomplished numerical outcomes conveniently reflect the productivity and competency of artificial neural network methods compared to customary approaches.

Aim. To develop a patient education algorithm used by medical workers. The leading method for studying this problem is a pedagogical experiment, as well as a statistical processing of quantitative research results. The relevance of this study is due to the need to address the priorities of the Healthcare national project, i.e. developing the skills of maintaining a healthy lifestyle among the population, changing behavior and attitudes towards their own health and the health of others. The introduction of educational programs for patients in combination with therapeutic measures contributes to the most complete rehabilitation. Forming partnerships and trust between patients and medical workers to prevent diseases requires the latter to have psychological and pedagogical knowledge and skills.Material and methods. The study was conducted on the basis of the City Outpatient Clinic № 7 (Kazan). The study was conducted in the Department of Internal Medicine on the daytime basis. The experiment involved 20 nurses, 15 doctors and 60 patients from these departments. The study was carried out in three following stages: at the first stage, the current state of the problem under study in pedagogical theory and practice was analyzed, a research methodology program was developed, and a survey of nurses at the medical institution was conducted. At the second stage, the theoretical concept of the study was adjusted, a training experiment was conducted with doctors and senior nurses, and a structural and functional model of pedagogical training of medical workers for patient education was introduced. At the third stage, analysis, generalization and systematization of the data obtained, and presentation of the research results were carried out.Results. The developed educational program for teaching patients, consisting of three stages (design, implementation, control of the educative process), and containing various methods of training, contributes to the formation of the psychological and pedagogical competence of medical workers. The specificity of the methodology is that the emphasis is on developing the ability to present information in an accessible form, and, most importantly, teach the patient and his family members specific techniques for self-monitoring, self-care and self-help.Conclusion. The study proved the effectiveness of the proposed algorithm for teaching patients by medical workers, which was confirmed by the development of patients’ skills and abilities in selfmonitoring and self-help. Patients' behavioral habits have changed, their motivation to improve their health has increased, and they have become responsible for their health.

We present a novel variational inequality model (VIM) to capture the complex real decision-making process in multi-tiered supply chain networks (MSCN) without strictly limiting the features of related functions. The VIM is formulated with the equilibrium conditions on links as the optimization goal and the flow conservation condition as the main constraints. We transform the VIM into a series of equivalent Non-Linear Programming Models (NLPMs) to solve. To address this challenge, we propose a novel population-based heuristic algorithm called the Multiscale Model Learning Algorithm (MMLA). The MMLA is inspired by the learning behavior of individuals in a group and can converge to an optimal equilibrium state of the MSCN. The MMLA has two key operations: zooming in on the search field and learning search in a learning stage. The excellent performers, called medalists, are imitated by other learners. With the increase in learning stages, the learning efficiency is improved, and the searching energy is concentrated in a more promising area. We employ sixteen benchmark optimization problems and two supply chain networks to demonstrate the effectiveness of the MMLA and the rationality of the equilibrium models. The results obtained by MMLA for the NLPM show that the MMLA can solve the equilibrium model effectively, and multiple optimal equilibrium states may exist for an MSCN. The flexibility of the NLPM makes it possible to consider more complicated decision-making mechanisms in the model.

In this paper, a novel fractional-order adaptive controller is presented for a class of nonlinear systems with unknown dynamics. The dynamics of the system is considered to be fully unknown. The multi-layer perceptron (MLP) neural network using restricted Boltzmann machine (RBMs) is employed for online dynamic identification. A deep learning method on the basis of contrastive divergence (CD) algorithm combined with the extended Kalman filter (EKF) is proposed for online optimisation. The proposed controller has two parts. The first part is a simple error feedback controller and the second one is the suggested DT2-FLS. The parameters of DT2-FLS are tuned such that a cost function of tracking error to be minimised and the closed-loop system to be stable. For the best knowledge of the authors, for the first time the tuning rules for the membership function and rule parameters of DT2-FLS are derived by error feedback learning method. The closed-loop stability is demonstrated with Lyapunov method and the well performance of the schemed controller is shown by applying on the induction motor and brushless DC motors. In addition to unknown dynamics, some disturbances are also considered such as abruptly changes in load torque and time-varying rotor resistance. Furthermore, the performance of the suggested scheme is compared with some popular controllers and FLSs. [ABSTRACT FROM AUTHOR]

In this paper, different learning methods based on Artificial Neural Networks (ANNs) are examined to replace the default speed controller for high-precision position control and drift attenuation in robotic manipulators. ANN learning methods including Levenberg–Marquardt and Bayesian Regression are implemented and compared using a UR5 robot with six degrees of freedom to improve trajectory tracking and minimize position error. Extensive simulation and experimental tests on the identification and control of the robot by means of the neural network controllers yield comparable results with respect to the classical controller, showing the feasibility of the proposed approach. [ABSTRACT FROM AUTHOR]

The use of intelligent methods in the mechatronics and robotics areas is a hot topic due to the need for the industry to switch to digital technologies. This paper presents an approach to machine learning of an electromechanical system control loop with an actuating mechanism. The real-time control system maintains the specified drive speed using a discrete proportional-integral controller implemented in MATLAB/Simulink. The system learning task is to maximize the actuator speed directly on the running plant while the controller parameters are adjusted automatically. Training is carried out using the simplex method and a random search algorithm. The results of the experiments are presented as illustrations of the operation of both algorithms. [ABSTRACT FROM AUTHOR]

Prevention of cervical cancer could be performed using Pap smear image analysis. This test screens pre-neoplastic changes in the cervical epithelial cells; accurate screening can reduce deaths caused by the disease. Pap smear test analysis is exhaustive and repetitive work performed visu-ally by a cytopathologist. This article proposes a workload-reducing algorithm for cervical cancer detection based on analysis of cell nuclei features within Pap smear images. We investigate eight traditional machine learning methods to perform a hierarchical classification. We propose a hierarchical classification methodology for computer-aided screening of cell lesions, which can recommend fields of view from the microscopy image based on the nuclei detection of cervical cells. We evaluate the performance of several algorithms against the Herlev and CRIC databases, using a varying number of classes during image classification. Results indicate that the hierarchical classification performed best when using Random Forest as the key classifier, particularly when compared with decision trees, k-NN, and the Ridge methods.

This ongoing work is vehemently dedicated to the investigation of a class of ordinary linear Volterra type integro-differential equations with fractional order in numerical mode. By replacing the unknown function by an appropriate multilayered feed-forward type neural structure, the fractional problem of such initial value is changed into a course of non-linear minimization equations, to some extent. Put differently, interest was sparked in structuring an optimized iterative first-order algorithm to estimate solutions for the origin fractional problem. On top of that, some computer simulation models exemplify the preciseness and well-functioning of the indicated iterative technique. The outstanding accomplished numerical outcomes conveniently reflect the productivity and competency of artificial neural network methods compared to customary approaches. [ABSTRACT FROM AUTHOR]

This paper presents a method for solving linear Fredholm integro-differential equations using a feedforward neural network based on Legendre polynomials. Firstly, the Legendre polynomials are used to approximate the unknown function and the kernel function in the equations. Secondly, The roots of Legendre polynomials are obtained by using Newton iteration method to obtain Gaussian integration points, and the obtained Gaussian integration points are used as the input nodes of the neural network, and the corresponding weight is learned by the gradient descent method to obtain an approximate solution. Finally, through the numerical value Case analysis to verify the effectiveness of the method. [ABSTRACT FROM AUTHOR]

The conventional tracking-learning-detection (TLD) algorithm is sensitive to illumination changes, clutter, significant changes of target shape between consecutive frames. In addition, low frame rate scenarios result in drift in object position or even missing the object. To solve these problems and enhance the tracking robustness, in this paper, TLD algorithm was extended in two folds. First, Kalman filter and Mean-shift algorithms were combined for the tracking part and second, co-training semi-supervised learning algorithm was used for the learning part of the conventional TLD structure. The Kalman filter estimates the position of the target in the next frame based on the previous positions of the target. This reduces tracking failure. On the other hands, the Mean-shift tracking algorithm is robust to rotation, partial occlusion and scale changing. In the learning part of TLD structure, two training tracking algorithms with two independent classifiers were run on the current frame simultaneously. Its structure makes data of both pools (color features and target templates) update by the results of other algorithm, in addition to the results of the corresponding algorithm in each of the tracking and detection algorithms. Therefore, classifiers can learn faster changing features of the target during the consecutive frames in online tracking process. Finally, the extended structure can solve the problem of lost object in LFR videos tracking and other similar challenges simultaneously. In terms of overlap ratio metric, comparing with conventional TLD and extended kernelized correlation filters (EKCF) algorithms, the success rate of our algorithm under various scenarios has increased by 161.03% and 255.82% respectively and under other scenarios, in terms of precision metric, it has increased by 18.46% and 1479.47%, respectively. Accordingly, comparative evaluations of the proposed method to other top state-of-the-art tracking algorithms under various scenarios present in TB-100 dataset, the superior performance of the proposed algorithm was demonstrated compared to other tracking algorithms in terms of tracking robustness, stability and real-time performance. [ABSTRACT FROM AUTHOR]

The use of quantitative techniques for the classification of population segments is a critical phase to evaluate their conditions. This information will serve as input for planning strategies to alleviate poverty. In this article, we present a model of vector support machines. Consequently, a sample of families residing in Cartagena de Indias is segmented, based on certain economic and sociodemographic variables. Analytical results confirm that most important factors are employment status, accessibility to public services and familiar income. In addition, it is corroborated that neighborhood conditions and monetary transfers have a low discriminatory power. [ABSTRACT FROM AUTHOR]

We propose an online learning model that efficiently teaches a defender's agent to learn the attacker's behavior while interacting in the cyber-world. This paper models the interaction between these two agents as a stochastic game with limited rationality. Because of this limited rationality, the proposed model helps the defender's agent learn the unknown communicator's behavior from the feedback obtained while interacting with it. Many models are built to solve the interaction between them by developing a state-oriented stochastic Markov game. However, such models fail due to the state explosion problem, and therefore, this paper discusses a model to solve this game, restricting it to a stateless stochastic game. The model is then compared to check the performance with different algorithms that solve stochastic games. The comparison between them shows that the proposed algorithm converges to an optimal strategy in a brief simulation time span. Finally, our model checks the performance with an existing technique that shows that the proposed algorithm chooses the correct strategy for around 91% of the simulation time compared to 73% of the simulation time by the existing algorithm. [ABSTRACT FROM AUTHOR]

Advances in technology have transformed and expanded the ways in which policing is run. One new manifestation is the mass acquisition and processing of private facial images via automatic facial recognition by the police: what we conceptualise as AFR-based policing. However, there is still a lack of clarity on the manner and extent to which this largely-unregulated technology is used by law enforcement agencies and on its impact on fundamental rights. Social understanding and involvement are still insufficient in the context of AFR technologies, which in turn affects social trust in and legitimacy and effectiveness of intelligent governance. This article delineates the function creep of this new concept, identifying the individual and collective harms it engenders. A technological, contextual perspective of the function creep of AFR in policing will evidence the comprehensive creep of training datasets and learning algorithms, which have by-passed an ignorant public. We thus argue individual harms to dignity, privacy and autonomy, combine to constitute a form of cultural harm, impacting directly on individuals and society as a whole. While recognising the limitations of what the law can achieve, we conclude by considering options for redress and the creation of an enhanced regulatory and oversight framework model, or Code of Conduct, as a means of encouraging cultural change from prevailing police indifference to enforcing respect for the human rights violations potentially engaged. The imperative will be to strengthen the top-level design and technical support of AFR policing, imbuing it with the values implicit in the rule of law, democratisation and scientisation-to enhance public confidence and trust in AFR social governance, and to promote civilised social governance in AFR policing. [ABSTRACT FROM AUTHOR]

With the increasingly prominent limitations of deep neural networks in practical applications,brain-like computing spiking neural networks with biological interpretability have become the focus of research.The uncertainty and complex diversity of application scenarios pose new challenges to researchers,requiring brain-like computing spiking neural networks with multi-scale architectures similar to biological brain organizations to realize the perception and decision-making function of multi-modal and uncertain information.This paper mainly introduces the multi-scale biological rational brain-like computing spiking neural network model and its learning algorithm for multi-modal information representation and uncertainty information perception,analyzing and discussing two key technical issues that the spiking neural network based on the interconnection of memristors can rea-lize multi-scale architecture brain-like computing,namely:the consistency problem of multi-modal and uncertain information with spike timing representation,and the computing fault-tolerant problem for the multi-scale spiking neural network with different learning algorithms.Finally,this paper analyzes and forecasts the further research direction of brain-like computing spiking neural network.

Heart failure is a very common disease, often a silent threat. It's also costly to treat and detect. There is also a steadily higher incidence rate of the disease at present. Although researchers have developed classification algorithms. Cardiovascular disease data were used by various ensemble learning methods, but the classification efficiency was not high enough due to the cumulative error that can occur from any weak learner effect and the accuracy of the vote-predicted class label. The objective of this research is the development of a new algorithm that improves the efficiency of the classification of patients with heart failure. This paper proposes Least Error Boosting (LEBoosting), a new algorithm that improves adaboost.m1's performance for higher classification accuracy. The learning algorithm finds the lowest error among various weak learners to be used to identify the lowest possible errors to update distribution to create the best final hypothesis in classification. Our trial will use the heart failure clinical records dataset, which contains 13 features of cardiac patients. Performance metrics are measured through precision, recall, f-measure, accuracy, and the ROC curve. Results from the experiment found that the proposed method had high performance compared to naïve bayes, k-NN,and decision tree, and outperformed other ensembles including bagging, logitBoost, LPBoost, and adaboost.m1, with an accuracy of 98.89%, and classified the capabilities of patients who died accurately as well compared to decision tree and bagging, which were completely indistinguishable. The findings of this study found that LEBoosting was able to maximize error reductions in the weak learner's training process from any weak learner to maximize the effectiveness of cardiology classifiers and to provide theoretical guidance to develop a model for analysis and prediction of heart disease. The novelty of this research is to improve original ensemble learning by finding the weak learner with the lowest error in order to update the best distribution to the final hypothesis, which will give LEBoosting the highest classification efficiency. Doi: 10.28991/ESJ-2023-07-01-010 Full Text: PDF

Copyright of Journal of South China University of Technology (Natural Science Edition) is the property of South China University of Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Introduction: Spiking Neural Networks (SNNs) are gaining significant traction in machine learning tasks where energy-efficiency is of utmost importance. Training such networks using the state-of-the-art back-propagation through time (BPTT) is, however, very time-consuming. Previous work employs an efficient GPU-accelerated backpropagation algorithm called SLAYER, which speeds up training considerably. SLAYER, however, does not take into account the neuron reset mechanism while computing the gradients, which we argue to be the source of numerical instability. To counteract this, SLAYER introduces a gradient scale hyper parameter across layers, which needs manual tuning. Methods: In this paper, we modify SLAYER and design an algorithm called EXODUS, that accounts for the neuron reset mechanism and applies the Implicit Function Theorem (IFT) to calculate the correct gradients (equivalent to those computed by BPTT). We furthermore eliminate the need for ad-hoc scaling of gradients, thus, reducing the training complexity tremendously. Results: We demonstrate, via computer simulations, that EXODUS is numerically stable and achieves comparable or better performance than SLAYER especially in various tasks with SNNs that rely on temporal features. [ABSTRACT FROM AUTHOR]

The conventional dermatology practice of performing noninvasive screening tests to detect skin diseases is a source of escapable diagnostic inaccuracies. Literature suggests that automated diagnosis is essential for improving diagnostic accuracies in medical fields such as dermatology, mammography, and colonography. Classification is an essential component of an assisted automation process that is rapidly gaining attention in the discipline of artificial intelligence for successful diagnosis, treatment, and recovery of patients. However, classifying skin lesions into multiple classes is challenging for most machine learning algorithms, especially for extremely imbalanced training datasets. This study proposes a novel ensemble deep learning algorithm based on the residual network with the next dimension and the dual path network with confidence preservation to improve the classification performance of skin lesions. The distributed computing paradigm was applied in the proposed algorithm to speed up the inference process by a factor of 0.25 for a faster classification of skin lesions. The algorithm was experimentally compared with 16 deep learning and 12 ensemble deep learning algorithms to establish its discriminating prowess. The experimental comparison was based on dermoscopic images congregated from the publicly available international skin imaging collaboration databases. We propitiously recorded up to 82.52% average sensitivity, 99.00% average specificity, 98.54% average balanced accuracy, and 92.84% multiclass accuracy without prior segmentation of skin lesions to outstrip numerous state-of-the-art deep learning algorithms investigated. [ABSTRACT FROM AUTHOR]

Chaos dynamics is an interesting nonlinear effect that can be observed in many chemical, electrical, and mechanical systems. The chaos phenomenon has many applications in various branches of engineering. On the other hand, the control of mobile robots to track unpredictable chaotic trajectories has a valuable application in many security problems and military missions. The main objective in this problem is to design a controller such that the robot tracks a desired chaotic path. In this paper, the concept of synchronization of chaotic systems is studied, and a new type-3 fuzzy system (T3FLS)-based controller is designed. The T3FLS is learned by some new adaptive rules. The new learning scheme of T3FLS helps to better stabilize and synchronize. The suggested controller has a better ability to cope with high-level uncertainties. Because, in addition to the fact that the T3FLSs have better ability in an uncertain environment, the designed compensator also improves the accuracy and robustness. Several simulations show better synchronization and control accuracy of the designed controller. [ABSTRACT FROM AUTHOR]

The size of datasets is growing exponentially as information technology advances, and it is becoming more and more crucial to provide efficient learning algorithms for neural networks to handle massive amounts of data. Due to their potential for handling huge datasets, feed-forward neural networks with random weights (FNNRWs) have drawn a lot of attention. In this paper, we introduced an efficient feed-forward neural network scheme (FNNS) for processing massive datasets with random weights. The FNNS divides large-scale data into subsets of the same size, and each subset derives the corresponding submodel. According to the activation function, the optimal range of input weights and biases is calculated. The input weight and biases are randomly generated in this range, and the iterative scheme is used to evaluate the output weight. The MNIST dataset was used as the basis for experiments. The experimental results demonstrate that the algorithm has a promising future in processing massive datasets. [ABSTRACT FROM AUTHOR]

In this paper, different learning methods based on Artificial Neural Networks (ANNs) are examined to replace the default speed controller for high-precision position control and drift attenuation in robotic manipulators. ANN learning methods including Levenberg–Marquardt and Bayesian Regression are implemented and compared using a UR5 robot with six degrees of freedom to improve trajectory tracking and minimize position error. Extensive simulation and experimental tests on the identification and control of the robot by means of the neural network controllers yield comparable results with respect to the classical controller, showing the feasibility of the proposed approach.

In this paper, we investigate a resource allocation problem for a Cellular Vehicle to Everything (C-V2X) network to improve energy efficiency of the system. To address this problem, self-organizing mechanisms are proposed for joint and disjoint subcarrier and power allocation procedures which are performed in a fully distributed manner. A multi-agent Q-learning algorithm is proposed for the joint power and subcarrier allocation. In addition, for the sake of simplicity, it is decoupled into two sub-problems: a subcarrier allocation sub-problem and a power allocation sub-problem. First, to allocate the subcarrier among users, a distributed Q-learning method is proposed. Then, given the optimal subcarriers, a dynamic power allocation mechanism is proposed where the problem is modeled as a non-cooperative game. To solve the problem, a no-regret learning algorithm is utilized. To evaluate the performance of the proposed approaches, other learning mechanisms are used which are presented in Fig. 8. Simulation results show the multi-agent joint Q-learning algorithm yields significant performance gains of up to about 11% and 18% in terms of energy efficiency compared to proposed disjoint mechanism and the third disjoint Q-learning mechanism for allocating the power and subcarrier to each user; however, the multi-agent joint Q-learning algorithm uses more memory than disjoint methods.

IntroductionSpiking Neural Networks (SNNs) are gaining significant traction in machine learning tasks where energy-efficiency is of utmost importance. Training such networks using the state-of-the-art back-propagation through time (BPTT) is, however, very time-consuming. Previous work employs an efficient GPU-accelerated backpropagation algorithm called SLAYER, which speeds up training considerably. SLAYER, however, does not take into account the neuron reset mechanism while computing the gradients, which we argue to be the source of numerical instability. To counteract this, SLAYER introduces a gradient scale hyper parameter across layers, which needs manual tuning.MethodsIn this paper, we modify SLAYER and design an algorithm called EXODUS, that accounts for the neuron reset mechanism and applies the Implicit Function Theorem (IFT) to calculate the correct gradients (equivalent to those computed by BPTT). We furthermore eliminate the need for ad-hoc scaling of gradients, thus, reducing the training complexity tremendously.ResultsWe demonstrate, via computer simulations, that EXODUS is numerically stable and achieves comparable or better performance than SLAYER especially in various tasks with SNNs that rely on temporal features.

It is known that the use of a multilayer perceptron with a traditional structure in solving real problems of image recognition and classification causes certain difficulties, in particular, associated with a large image dimension (this significantly increases the number of neurons and synaptic connections in neural connections of the network and, therefore, significantly increases the training sample and training time). In addition, the topology of the input data is ignored. The components of the input layer can be presented in any order, regardless of the learning goal. However, images have a strict two-dimensional structure, in which there is a relationship between neighboring pixels in space. These shortcomings are deprived of the so-called convolutional neural networks, which are a special class of multilayer perceptrons specifically designed to recognize two-dimensional surfaces with a high degree of invariance to scaling, displacement, rotation, angle change, and other spatial transformations. This article discusses the problem of practical implementation of handwritten digit recognition based on convolutional neural networks (CNN). The CNN architecture is presented, for which it is recommended to use cross-entropy as the learning loss function, and the Softmax function as the activation function of the last CNN layer. It is also proposed to use a modification of the well-known error backpropagation algorithm to implement the CNN learning algorithm. To do this, the article presents the main ratios for errors at each level. Ref. 14, pic.2.

Good feature representation is the chief requirement for improving Language Identification (LID) system recognition performance. In this work LID system for Indian languages is proposed based on unsupervised feature learning utilizing Deep Belief Network (DBN). The proposed methodology is implemented in two parts. The first phase of this work is based on extracting MFCC features combined with SDC hybrid features. The resultant hybrid features are further stacked to Deep Belief Network (DBN). The second phase of the proposed work is investigating the performance of various Feed forward back propagation neural network models for classification using different training algorithms. Effect of combining different activation functions and varying the hidden neurons is also investigated The performance of the resultant models is evaluated on the basis of some performance metrics such as the epochs, training time, Mean Square Error, Regression and Mean Absolute Percentage Error. Results indicate that optimal performance is achieved in model trained with Levenberg Marquardt (LM) training algorithm. The activation functions used in the hidden and output layer are "tansig" and "purelin". Similarly, the effect of varying the number of neurons in the hidden layer is not significant in improving the performance of the derived models. FFBPNN models trained with PL and TS activation functions gave best performance indices. A user defined language database in four different languages Hindi, English, Tamil and Malayalam is used for this work. [ABSTRACT FROM AUTHOR]

An algorithm for training a multiagent control system for an electrical-engineering facility of an oil- and gas-production enterprise with distributed generation has been developed using simulation tools of the RastrWin3 and LabVIEW software packages. The main purpose of the learning algorithm is to provide self-tuning of a multiagent control system in the case in which the system topology and/or principles of interaction between elements change. To assess the operability of the algorithm, a model of the electrical-engineering facility of the Southern Dome of the Yurchuk deposit (Perm krai) has been developed. The multiagent control system is designed to maintain energy balance and enables creating a virtual environment (a digital twin of the electrical-engineering facility of the deposit), which simulates scenarios for training agents. The multiagent control system models the current situations as a Markov process and finds steering scenarios that provide minimal deviations from normal energy-balance indicators. If the operation mode is modeled without using reinforcement learning during the test interval, the energy balance in some scenarios goes beyond the established limits of adequacy. [ABSTRACT FROM AUTHOR]