Search

In this paper, a set of machine learning (ML) tools is applied to estimate the water saturation of shallow unconsolidated sediments at the Bátaapáti site in Hungary. Water saturation is directly calculated from the first factor extracted from a set of direct push logs by factor analysis. The dataset observed by engineering geophysical sounding tools as special variants of direct-push probes contains data from a total of 12 shallow penetration holes. Both one- and two-dimensional applications of the suggested method are presented. To improve the performance of factor analysis, particle swarm optimization (PSO) is applied to give a globally optimized estimate for the factor scores. Furthermore, by a hyperparameter estimation approach, some control parameters of the utilized PSO algorithm are automatically estimated by simulated annealing (SA) to ensure the convergence of the procedure. The result of the suggested ML-based log analysis method is compared and verified by an independent inversion estimate. The study shows that the PSO-based factor analysis aided by hyperparameter estimation provides reliable in situ estimates of water saturation, which may improve the solution of environmental end engineering problems in shallow unconsolidated heterogeneous formations. [ABSTRACT FROM AUTHOR]

This paper proposes a hybrid robust-adaptive learning-based control scheme based on Approximate Dynamic Programming (ADP) for the tracking control of autonomous ship maneuvering. We adopt a Time-Delay Control (TDC) approach, which is known as a simple, practical, model free and roughly robust strategy, combined with an Actor-Critic Approximate Dynamic Programming (ACADP) algorithm as an adaptive part in the proposed hybrid control algorithm. Based on this integration, Actor-Critic Time-Delay Control (AC-TDC) is proposed. It offers a high-performance robust-adaptive control approach for path following of autonomous ships under deterministic and stochastic disturbances induced by the winds, waves, and ocean currents. Computer simulations have been conducted under two different conditions in terms of the deterministic and stochastic disturbances and all simulation results indicate an acceptable performance in tracking of paths for the proposed control algorithm in comparison with the conventional TDC approach. [ABSTRACT FROM AUTHOR]

There has been a recent growth of interest in software engineering community to use grammatical inference, aka automaton learning, in software engineering applications. This is primarily due to the reason that capacity of underlying hardware resources has improved significantly over the last years; which has enabled the use of this approach beyond the toy examples in a greater frequency. In this paper, we present a new grammatical inference algorithm, DKL, to learn deterministic Kripke structures. The DKL is only the second of its kind for incremental learning of deterministic Kripke structures. Earlier, IKL algorithm was introduced for learning deterministic Kripke structures but it often constructs a hypothesis much larger in state size than the target Kripke structure. This problem is known as state-space explosion and it primarily occurs due to the sub-direct product construction used in the IKL design, which in turn affects time efficiency of IKL; especially when it is used for practical applications of software engineering. The DKL algorithm is designed to resolve the problem of state-space explosion. We give a proof of correctness and termination of the DKL algorithm. We also compared the performance of DKL with IKL by implementing an evaluation framework. Our results show that DKL is more efficient in terms of time in reaching the target automaton than the IKL algorithm and is not prone to the problem of state-space explosion. [ABSTRACT FROM AUTHOR]

Fuzzy support vector machine and their variants are prominent classification techniques that reduce the adverse effects of noises and outliers as compare to classical support vector machine. However, in these methods, only the distance between the training pattern and the class center is considered, and hence, the edge support vectors cannot be distinguished from noises and outliers properly. These limitations are overcome by intuitionistic fuzzy-based support vector machine which allocate two parameters (membership and non-membership degrees) to each pattern of a dataset and hence define score number based on the importance of a pattern. In this paper, an intuitionistic fuzzy proximal support vector machine for multi-category classification problems is proposed. The method significantly reduces the impacts of noises and outliers present in the dataset by assigning the intuitionistic fuzzy score function to each training point based on its location and surroundings. Moreover, the method is computationally efficient as the robust classifiers are obtained by solving the system of linear equations instead of large size quadratic programming problems. In the proposed method, using polynomial and Gaussian kernels, the hyperplanes are also developed in the feature space. The geometrical advantages of the suggested method over the existing techniques are ascertained using the simulated two-dimensional artificial dataset having three target classes. Further, extensive experimental studies on ten UCI benchmark datasets have been performed which demonstrate that the proposed algorithm predicts more precisely about future data as comparison to some well-established algorithms. Figures are also illustrated by varying different parameters involved in the model which confirms the performance of the method over the existing algorithms. Moreover, the analysis of the predictive behavior of the proposed approach is done using the Friedman test, a nonparametric alternative to the analysis of variance test at 5 % significance level. Further, the proposed method has also been applied to image classification and gesture-phase segmentation problems which confirms the efficiency and handling capabilities of the proposed algorithm in practical applications. [ABSTRACT FROM AUTHOR]

Prediction of rockburst in underground engineering is of significance as it is close to the safety of support structures, personnel, and working environments. To improve the accuracy of rockburst classification, this paper proposed an ensemble classifier RF-FA model in which the random forest classifier (RF) and firefly algorithm (FA) were combined to achieve the optimum performance on rockburst prediction. Five key parameters of surrounding rock, i.e., the depth H, the maximum tangential stress σθ, the uniaxial compressive strength σc, the tensile strength σt, and the elastic energy index Wet, are selected as input variables while the rockburst intensity including none, light, moderate, and strong classes was chosen as output. A total of 279 cases worldwide were collected and used for train and test the proposed RF-FA model. The results showed that the FA can optimize the hyperparameters of RF efficiently and the optimum model exhibited a high performance on rockburst data from the independent test set and new engineering projects. The feature importance obtained by the ensemble RF-FA model indicated that the elastic energy index plays the most important role in rockburst. Besides, the proposed model showed much better accuracy on rockburst classification compared with previously existing RF models and empirical criteria, which means it is a useful and robust tool for rockburst prediction. [ABSTRACT FROM AUTHOR]

Spike trains (STs) induced by external stimuli are complex and challenging to decode the embedded spatiotemporal information. Although various methods have been developed to characterize STs, few applications have been reported in tactile discrimination applications. In this paper, a neurocomputing method based on reference spike train (RST) is proposed to establish a neural computation scheme, on which existing ST metrics could be fed into various traditional models directly. Moreover, existing metrics in the field of statistics and vector measurement are introduced together to extract more discriminative features. With the binning technique and feature selection algorithm applied, the neural computation scheme is aimed at taking advantage of as maximal as possible information contained in tactile signals. Based on our designed artificial fingertip, the effect is validated by improving the recognition accuracy from 77.6% to 83.4% when it is applied to the discrimination of eight roughness surfaces. Furthermore, properties of RST, such as spike intervals and distributions, are evaluated and it is found that RSTs with uniform distribution perform the best for tactile discrimination. [ABSTRACT FROM AUTHOR]

In many machine learning applications and algorithms, the algorithm performance and accuracy are highly dependent on the metric used to measure the distance between different samples. Therefore, learning a distance metric specific to the data can improve these algorithms' performance. This paper proposes an unrestricted deep metric learning framework based on neural networks' interaction for learning metrics in latent space. The proposed method is inspired by generative neural nets (GANs), in which two neural nets are working together to learn true data distribution. In our method, one network plays the role of a supervisor for another network, a feature learning auto-encoder. Its task is to learn transformation to latent space in which data have more meaningful distance and separability. i.e., the supervisor gets the output of the auto-encoder and sends feedback to modify its weights. They interact with each other interleavingly. Several experiments were conducted on four datasets, such as MNIST, GISETTE, Winnipeg Cropland Classification (WCC), and swarm behavior, from different application domains, to evaluate the proposed method's performance. The results show that we can force auto-encoder to learn label information to project data into a latent space with better separability by using our approach. In addition to better class discrimination, the proposed method is far faster than normal auto-encoders during feature learning and has much less training time in the classification phase. [ABSTRACT FROM AUTHOR]

It may not be easy to implement complex AI models in edge devices without strong computing capacity (e.g. GPU). The Universal Approximation Theorem states that a shallow neural network (SNN) can represent any nonlinear function. In this paper, we focus on the learnability and robustness of SNNs, obtained by a greedy tight force heuristic algorithm [a Performance Driven Back-Propagation (PDBP)] and a loose force meta-heuristic algorithm [a variant of particle swarm optimization (VPSO)]. From the engineering prospective, all sensors are well justified for a specific task. Hence, all sensor readings should be strongly correlated to the target, and the structure of an SNN should depend on the dimensions of a problem space. The key findings of the research are summarized as follows: (1) The number of hidden neurons of an SNN depends on the nonlinearity of the training data, and the number of hidden neurons up to the dimension number of a problem space could be enough; (2) The learnability of SNNs, produced by error-driven PDBP, is always better than that of SNNs, optimized by error-driven VPSOs; (3) The performances of SNNs, obtained by PDBPs and VPSOs, do not change much for different training rates; and (4) Comparing with other classic machine learning algorithms, such as C4.5, NB and NN in literature, the SNNs, obtained by accuracy-driven PDBPs, win for all tested data sets, and the improvement percentage is up to 32.86%. Hence, the research could provide valued guidance for the implementation of edge intelligence. [ABSTRACT FROM AUTHOR]

Accurate demand prediction of bike-sharing is an important prerequisite to reducing the cost of scheduling and improving the user satisfaction. However, it is a challenging issue due to stochasticity and non-linearity in bike-sharing systems. In this paper, a model called pseudo-double hidden layer feedforward neural networks is proposed to approximately predict actual demands of bike-sharing. Specifically, to overcome limitations in traditional back-propagation learning process, an algorithm, an extreme learning machine with improved particle swarm optimization, is designed to construct learning rules in neural networks. The performance is verified by comparing with other learning algorithms on the dataset of Streeter Dr bike-sharing station in Chicago. [ABSTRACT FROM AUTHOR]

Solar flares are one of the most important sources of disastrous space weather events, leading to negative effects on spacecrafts and living organisms. It is very important to predict solar flares to minimize the potential losses. In this paper, we use three different machine learning algorithms: K-Nearest Neighbors (KNN), Random Forest (RF), and XGBoost (XGB) to predict the total flare index T flare and the maximum flare index M flare of an active region (AR) within the subsequent of 24, 48, and 72 hrs. First, we selected 54514 vector magnetograms of 129 ARs on the visible solar hemisphere in solar cycle 24 whose maximum sunspot groups' area was larger than 400 μh. Then the following four magnetic parameters of each magnetogram were calculated: 1) the total magnetic flux | Φ tot | , 2) the total photospheric free magnetic energy density E free , 3) the gradient-weighted integral length of the neutral line with horizontal magnetic gradient of line-of-sight magnetic field larger than 0.1 G km − 1 ( WL SG ), and 4) the area with magnetic shear angle larger than 40 ∘ ( A Ψ ), as well as T flare and M flare corresponding to each magnetogram. Afterward, we split samples randomly into training (85% of the whole data) and testing (15%) data sets. After hyperparameter tuning and model construction we found that RF is an optimal algorithm for the prediction task and that the coefficients of determination ( R 2 ) of test data set via the majority of RF models are beyond 0.97. In addition, the feature importance of RF and XGB models indicates that | Φ tot | and E free are two optimal parameters to predict both T flare and M flare , and | Φ tot | and E free are the best parameters for M flare and T flare , respectively. [ABSTRACT FROM AUTHOR]

The efficient use of resources in wireless communications has always been a major issue. In the Internet of Things (IoT), the energy resource becomes more critical. The transmission policy with the aid of a coordinator is not a viable solution in an IoT network, since a node should report its state to the coordinator for scheduling and it causes serious signaling overhead. Machine learning algorithms can provide the optimal distributed transmission mechanism with little overhead. A node can learn by itself by utilizing the machine learning algorithm and make the optimal transmission decision on its own. In this paper, we propose a novel learning Medium Access Control (MAC) protocol with learning nodes. Nodes learn the optimal transmission policy, i.e., minimizing the data and energy queue levels, using the Q-learning algorithm. The performance evaluation shows that the proposed scheme enhances the queue states and throughput. [ABSTRACT FROM AUTHOR]

Currently, Iris detection is considered as a significant module for robust biometric systems and high-speed applications such as eye tracking. Most iris segmentation models are based on machine learning algorithms or geometric methods. In this paper, we use an elliptical Hough transform to firstly detect the shape of the palpebral fissure. Then, a correlation-based circular Hough transform (we named it CCHT) is proposed to extract iris from the surrounding structures. One of the advantages of the proposed method is its ability to determine the closed-eye images, in order to remove these images in the process of eye tracking procedure. Moreover, the algorithm is simple and fast which make it suitable for on-line eye tracking. Experimental results on UBIRIS, which contains some defocused and eyelid-occluded images as non-ideal and noisy frames, indicate that the proposed method is efficient and much faster, in comparison with the previous approaches and encouraging improved accuracy on iris detection. [ABSTRACT FROM AUTHOR]

Computer-aided diagnosis system is becoming a more and more important tool in clinical treatment, which can provide a verification of the doctors' decisions. In this paper, we proposed a novel abnormal brain detection method for magnetic resonance image. Firstly, a pre-trained AlexNet was modified with batch normalization layers and trained on our brain images. Then, the last several layers were replaced with an extreme learning machine. A searching method was proposed to find the best number of layers to be replaced. Finally, the extreme learning machine was optimized by chaotic bat algorithm to obtain better classification performance. Experiment results based on 5 × hold-out validation revealed that our method achieved state-of-the-art performance. [ABSTRACT FROM AUTHOR]

The paper observes a similarity between the stochastic optimal control of discrete dynamical systems and the learning multilayer neural networks. It focuses on contemporary deep networks with nonconvex nonsmooth loss and activation functions. The machine learning problems are treated as nonconvex nonsmooth stochastic optimization problems. As a model of nonsmooth nonconvex dependences, the so-called generalized-differentiable functions are used. The backpropagation method for calculating stochastic generalized gradients of the learning quality functional for such systems is substantiated basing on Hamilton–Pontryagin formalism. Stochastic generalized gradient learning algorithms are extended for training nonconvex nonsmooth neural networks. The performance of a stochastic generalized gradient algorithm is illustrated by the linear multiclass classification problem. [ABSTRACT FROM AUTHOR]

The unique properties of terahertz (THz) spectroscopy show a great potential for security and defense applications such as safe screening of persons and objects. However, a successful implementation of THz screening systems requires a development of reliable and efficient identification algorithms. Dimensionality reduction (DR) methods aim to reduce the dimensionality of the multivariate data and are therefore commonly used as a preprocessing step for classification algorithms and as an analytical tool allowing data visualization. In this paper, we compare the use of unsupervised and supervised DR methods for analysis and classification of THz reflection spectra based on their most widespread linear representatives, namely principal component analysis and linear discriminant analysis, respectively. To this end, both methods were applied to more than 5000 THz reflection spectra acquired from six active materials mixed at three different concentrations with polyethylene and measured at various humidity conditions. While considering scenarios with different levels of complexity, we found that the supervised approach provides better results because it enables efficient grouping despite intra-class variability. Furthermore, we showed that manipulating labels introduced into the supervised DR algorithm allows conditioning the data for a desired classification task such as security screening. Presented classification results show that simple machine learning algorithms are sufficient for highly accurate classification (>98.6%) of THz spectra, which will be suitable for many real-life applications of THz spectroscopy based on material identification. [ABSTRACT FROM AUTHOR]

Although tendon-driven continuum manipulators have been extensively researched, how to realize tip contact force sensing in a more general and efficient way without increasing the diameter is still a challenge. Rather than use a complex modeling approach, this paper proposes a general tip contact force-sensing method based on a recurrent neural network that takes the tendons' position and tension as the input of a recurrent neural network and the tip contact force of the continuum manipulator as the output and fits this static model by means of machine learning so that it may be used as a real-time contact force estimator. We also designed and built a corresponding three-degree-of-freedom contact force data acquisition platform based on the structure of a continuum manipulator designed in our previous studies. After obtaining training data, we built and compared the performances of a multi-layer perceptron-based contact force estimator as a baseline and three typical recurrent neural network-based contact force estimators through TensorFlow framework to verify the feasibility of this method. We also proposed a manually decoupled sub-estimators algorithm and evaluated the advantages and disadvantages of those two methods. [ABSTRACT FROM AUTHOR]

Local feature descriptor plays a significant role in texture classification. However, in the traditional local binary pattern method, image pixels are converted into a binary pattern based on the relationship between center and neighborhood pixels. This paper introduces a novel feature extraction method named local neighbourhood edge responsive binary pattern to extract and categorize reliable texture features from images for achieving an automated classification of different types of ocean bottom sediments. Initially, the basic magnitude variance pixel values are derived depending on an odd and even pixel value of a 3 × 3 image patch. Further, the edge information is extracted using the local directional pattern method from all images. The edge response of the image is obtained using a kirsch mask in all eight directions. The encoding condition is then applied to both the local intensity and the edge information to create a unique descriptor value. Finally, a new learning algorithm called GMJAYA-ELM combines the Gaussian mutated JAYA (GMJAYA) with an extreme learning machine (ELM) for texture classification. The GMJAYA is used to optimize the input weights and hidden biases of single-hidden-layer feed-forward neural networks. Experimental findings indicate that the suggested method yields better accuracy and sensitivity efficiency among various groups. The proposed algorithm is tested by contrasting outcomes with standard learning schemes such as PSO-ELM, GA-ELM, ABC-ELM, Birdswarm-ELM, and JAYA-ELM, suggesting the dominance of GMJAYA-ELM. [ABSTRACT FROM AUTHOR]

With the development of quantum computation theory, some researchers further apply it to improve the efficiency of classical machine learning algorithms. Based on physical graph state, an efficient quantum version of classifier is proposed in this paper. Different from existing classical methods, realizing the proposed scheme is an entangling-subgraphs process of physical system to build the classifier by using the graph states. Resort to the efficiency of graph state, the quantum algorithm is more efficient to big data than the classical ones. [ABSTRACT FROM AUTHOR]

In this paper, we proposed a novel regularization and variable selection algorithm called Liu–Lasso extreme learning machine (LL-ELM) in order to deal with the ELM's drawbacks like instability, poor generalizability and underfitting or overfitting due to the selection of inappropriate hidden layer nodes. Liu estimator, which is a statistically biased estimator, is considered in the learning phase of the proposed algorithm with Lasso regression approach. The proposed algorithm is compared with the conventional ELM and its variants including ELM forms based on Liu estimator (Liu-ELM), L 1 -norm (Lasso-ELM), L 2 -norm (Ridge-ELM) and elastic net (Enet-ELM). Convenient selection methods for the determination of tuning parameters for each algorithm have been used in comparisons. The results show that there always exists a d value such that LL-ELM overperforms either Lasso-ELM or Enet-ELM in terms of learning and generalization performance. This improvement in LL-Lasso over Lasso-ELM and Enet-ELM in the sense of testing root mean square error varies up to 27 % depending on the proposed d selection methods. Consequently, LL-ELM can be considered as a competitive algorithm for both regressional and classification tasks because of easy integrability property. [ABSTRACT FROM AUTHOR]

Heat exchangers are applicable in different industries and technologies, and their performance is influenced by different parameters. In addition to experimental and time-consuming computational approaches, intelligent methods can be used for the investigation of heat exchanger performance due to their abilities in accurate prediction and relatively fast performance. The accuracy and applicability of machine learning methods, mainly based on intelligent techniques, in modeling and forecasting the performance of heat exchangers are dependent on some factors including architecture of algorithm, inputs of the model, and complexity of the system. Owing to the aforementioned facts, it would be crucial to consider the influential factors in the proposed mode to produce models with the greatest accuracy. In this work, different applications of intelligent methods in performance modeling heat exchangers are reviewed, and the key outcomes of the reviewed works are represented. Moreover, the items influencing the performance of these methods are investigated. In the final stage of the current paper, some ideas are recommended for future works in the relevant fields. [ABSTRACT FROM AUTHOR]

In recent decades, researches on optimizing the parameter of the artificial neural network (ANN) model has attracted significant attention from researchers. Hybridization of superior algorithms helps improving optimization performance and capable of solving complex applications. As a traditional gradient-based learning algorithm, ANN suffers from a slow learning rate and is easily trapped in local minima when training techniques such as gradient descent (GD) and back-propagation (BP) algorithm are used. The characteristics of randomization and selection of the best or near-optimal solution of metaheuristic algorithm provide an effective and robust solution; therefore, it has always been used in training of ANN to improve and overcome the above problems. New metaheuristic algorithms are proposed every year. Therefore, the review of its latest developments is essential. This article attempts to summarize the metaheuristic algorithms which have been proposed from the year 1975 to 2020 from various journals, conferences, technical papers, and books. The comparison of the popularity of the metaheuristic algorithm is presented in two time frames, such as algorithms proposed in the recent 20 years and those proposed earlier. Then, some of the popular metaheuristic algorithms and their working principle are reviewed. This article further categorizes the latest metaheuristic search algorithm in the literature to indicate their efficiency in training ANN for various industry applications. More and more researchers tend to develop new hybrid optimization tools by combining two or more metaheuristic algorithms to optimize the training parameters of ANN. Generally, the algorithm's optimal performance must be able to achieve a fine balance of their exploration and exploitation characteristics. Hence, this article tries to compare and summarize the properties of various metaheuristic algorithms in terms of their convergence rate and the ability to avoid the local minima. This information is useful for researchers working on algorithm hybridization by providing a good understanding of the convergence rate and the ability to find a global optimum. [ABSTRACT FROM AUTHOR]

Feature selection is one of the main steps in preprocessing data in machine learning, and its goal is to reduce features by removing additional and noisy features. Feature selection methods and feature reduction in a dataset must consider the accuracy of the classifying algorithms. Meta-heuristic algorithms serve as the most successful and promising methods to solve this problem. Symbiotic Organisms Search (SOS) is one of the most successful meta-heuristic algorithms inspired by organisms' interaction in nature called mutualism, commensalism, and parasitism. In this paper, three SOS-based binary approaches are offered to solve the feature selection problem. In the first and second approaches, several S-shaped transfer functions and several Chaotic Tent Function-based V-shaped transfer functions called BSOSST and BSOSVT are used to make the binary SOS (BSOS). In the third approach, an advanced BSOS based on changing SOS and the chaotic Tent function operators called EBCSOS is provided. The EBCSOS algorithm uses the chaotic Tent function and the Gaussian mutation to increase usefulness and exploration. Moreover, two new operators, i.e., BMPT and BCPT, are suggested to make the commensalism and mutualism stage binary based on a chaotic function to solve the feature selection problem. Finally, the proposed BSOSST and BSOSVT methods and the advanced version of EBCSOS were implemented on 25 datasets than the basic algorithm's binary meta-heuristic algorithms. Various experiments demonstrated that the proposed EBCSOS algorithm outperformed other methods in terms of several features and accuracy. To further confirm the proposed EBCSOS algorithm, the problem of detecting spam E-mails was applied, with the results of this experiment indicating that the proposed EBCSOS algorithm significantly improved the accuracy and speed of all categories in detecting spam E-mails. [ABSTRACT FROM AUTHOR]

With growing heterogeneity and complexity in applications, demand to design an energy-efficient and fast computing system in multi-core architecture has heightened. This paper presents a regression-based dynamic voltage frequency scaling model which studies and utilizes workload characteristics to obtain optimal voltage–frequency (v–f) settings. The proposed framework leverages the workload profile information together with power constraints to compute the best-suited voltage–frequency (v–f) settings to (a) maintain global power budget at chip-level, (b) maximize performance while enforcing power constraints at the per-core level. The presented algorithm works in conjunction with the workload characterizer and senses change in application requirements and apply the knowledge to select the next setting for the core. Our results when compared with two state-of-the-art algorithms MaxBIPS and TPEq achieve the average power reduction of 33% and 25% respectively across 32-core architecture for PARSEC benchmarks. [ABSTRACT FROM AUTHOR]

We explore a new approach for training neural networks where all loss functions are replaced by hard constraints. The same approach is very successful in phase retrieval, where signals are reconstructed from magnitude constraints and general characteristics (sparsity, support, etc.). Instead of taking gradient steps, the optimizer in the constraint based approach, called relaxed–reflect–reflect (RRR), derives its steps from projections to local constraints. In neural networks one such projection makes the minimal modification to the inputs x, the associated weights w, and the pre-activation value y at each neuron, to satisfy the equation x ⋅ w = y . These projections, along with a host of other local projections (constraining pre- and post-activations, etc.) can be partitioned into two sets such that all the projections in each set can be applied concurrently—across the network and across all data in the training batch. This partitioning into two sets is analogous to the situation in phase retrieval and the setting for which the general purpose RRR optimizer was designed. Owing to the novelty of the method, this paper also serves as a self-contained tutorial. Starting with a single-layer network that performs nonnegative matrix factorization, and concluding with a generative model comprising an autoencoder and classifier, all applications and their implementations by projections are described in complete detail. Although the new approach has the potential to extend the scope of neural networks (e.g. by defining activation not through functions but constraint sets), most of the featured models are standard to allow comparison with stochastic gradient descent. [ABSTRACT FROM AUTHOR]

Internet is a new type of information exchange tool that has developed with the times. Now, it has been integrated into all aspects of our study and life. At the same time, in the era of the rise of social media, there are more and more platforms on the Internet. A variety of critical texts have also exploded in these platforms. In these opinions and comments, the subjective opinions of the presenters are included, and the emotional tendencies of the commenters are expressed. Nowadays, subjective text information resources are huge. One of the major problems to be solved by management objects facing information management is how to manage them effectively so that users can quickly and accurately find the required information. Therefore, classifying these text tendencies and mining the potential value in the text has broad application prospects. Based on the text granularity and processing efficiency, this paper conducts multi-granularity emotional block partitioning on network texts and compares the sentiment analysis under different granularities. Furthermore, a random subspace integrated learning text sentiment classification method based on BPSO (binary particle swarm optimization) is proposed to analyze text orientation. By simulating news site comments and e-commerce website reviews such as Taobao, the convergence analysis of BPSO in the optimization of the number of base classifiers shows that the BPSO algorithm can be applied to the random subspace method well, and the accuracy of the classification results is high. [ABSTRACT FROM AUTHOR]

Efficient face recognition can realize fast and accurate face recognition and make it widely used in essential fields such as human–computer interaction and access control. At present, there are many face recognition methods whose recognition rate can reach high accuracy, but the training of the model and the recognition of samples take much time, which leads to insufficient real-time performance. This paper designs a fusion facial semantic feature (FFSF) and an incremental learning mechanism (ILM) for efficient face recognition. FFSF feature is a fusion of facial contour features and facial semantic component features, which can extract contour features and interior features of facial organs (eyes, mouth, nose, and eyebrow) according to facial organs' position. FFSF features can ensure that the extracted features are concentrated in the face's most discriminative region, making the extracted features have good discriminative characteristics. Then, we use a clustering algorithm to construct a hierarchical incremental learning tree (HIL-Tree) with a hierarchical structure and use the HIL-Tree to implement the ILM. ILM achieves fast and accurate sample classification by retrieving the nodes in HIL-Tree, and the training samples can be directly added to the HIL-Tree by retrieval instead of rebuilding the HIL-Tree during the training process. Extensive experiments on several public data sets demonstrate the proposed efficient face recognition method's excellent accuracy and efficiency. [ABSTRACT FROM AUTHOR]

Formal concept analysis (FCA) is an unsupervised machine learning technique used for knowledge discovery and representation. A major task in FCA is the enumeration of the implications to construct the implication base. Even though there are many efficient classical and parallel algorithms proposed for constructing the implication base, the existing algorithms are not well suited for large formal contexts because of their architectural complexity. All the existing works use either stem-base or proper-premise approach to find the implication base in exponential time. Hence, we introduce a distributed algorithm to find the implication base quickly in larger datasets in polynomial time. In this paper, we propose a scalable algorithm to find the implication base using machine learning technique FP-growth, big data processing framework Apache Spark and executed on large formal contexts. Extensive experiments on the real-world datasets show that the proposed algorithm has an improved gain in performance metrics such as execution time, CPU and memory usage. The statistical validations on the experimental results prove that the proposed algorithm has the better potential to find the implication base. [ABSTRACT FROM AUTHOR]

Machine learning (ML) algorithms, such as neural networks, random forest, and more recent deep learning, are illustrating their utility for waste recycling. The increasing computational power of ML makes waste generation prediction, even at municipal level, possible with satisfying accuracy. ML is so critical and efficient and yet it is severely under-researched in recycling business. Also, the ML application in the recycling business is still a niche area judged by the limitations in its literature sources, the research domains, the ML algorithms' use and benefits involved or reported in the literature. To unlock the value of ML in recycling business, this paper reviewed 51 related articles systematically and presents the current obstacles and future directions in applying ML to waste recycling industries. [ABSTRACT FROM AUTHOR]

Classification is one of the foremost machine learning tasks in this modern era. Neural Network (NN) is one of the powerful classification techniques. NN can achieve high classification accuracy on highly imbalanced and complex datasets, but lacks in explanation of its reasoning process which limits its applicability in various domains which require transparent decision along with good accuracy. There are some techniques which extract rules from NN and make it transparent; however, attribute pruning, rule pruning and class overlap algorithms are not sufficiently effective. Therefore, this paper proposes a rule extraction algorithm, called Transparent Rule Extraction using Neural Network (TRENN) to convert NN into white box with greater emphasis on attribute pruning and rule pruning. The proposed TRENN is a pedagogical approach and an extension of one of the existing algorithms named Rule Extraction from Neural Network using Classified and Misclassified data (RxNCM). The proposed TRENN extends the RxNCM with sequential floating backward search for feature and rule selection to improve the comprehensibility of the generated rules. Besides, the proposed TRENN uses probabilistic approach for the treatment of class overlapping problem in the rule updating phase instead of reclassification used in RxNCM where the overlap may persist. Experiments are conducted with eight real datasets collected from the UCI repository. Performance of the TRENN is measured with Precision, Recall, FP-Rate, F-measure, and local and global comprehensibility. It is observed from the experimental results that TRENN performs better than Re-RX, RxNCM and RxREN. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) is one of the main drivers of what has been described as the "Fourth Industrial Revolution", as well as the most innovative technology developed to date. It is a pervasive transformative innovation, which needs a new approach. In 2017, the European Parliament introduced the notion of the "electronic person", which sparked huge debates in philosophical, legal, technological, and other academic settings. The issues related to AI should be examined from an interdisciplinary perspective. In this paper, we examine this legal innovation—that has been proposed by the European Parliament—from not only legal but also technological points of view. In the first section, we define AI and analyse its main characteristics. We argue that, from a technical perspective, it appears premature and probably inappropriate to introduce AI personhood now. In the second section, justifications for the European Parliament's proposals are explored in contrast with the opposing arguments that have been presented. As the existing mechanisms of liability could be insufficient in scenarios where AI systems cause harm, especially when algorithms of AI learn and evolve on their own, there is a need to depart from traditional liability theories. [ABSTRACT FROM AUTHOR]

The stochastic Frank–Wolfe method has recently attracted much general interest in the context of optimization for statistical and machine learning due to its ability to work with a more general feasible region. However, there has been a complexity gap in the dependence on the optimality tolerance ε in the guaranteed convergence rate for stochastic Frank–Wolfe compared to its deterministic counterpart. In this work, we present a new generalized stochastic Frank–Wolfe method which closes this gap for the class of structured optimization problems encountered in statistical and machine learning characterized by empirical loss minimization with a certain type of "linear prediction" property (formally defined in the paper), which is typically present in loss minimization problems in practice. Our method also introduces the notion of a "substitute gradient" that is a not-necessarily-unbiased estimate of the gradient. We show that our new method is equivalent to a particular randomized coordinate mirror descent algorithm applied to the dual problem, which in turn provides a new interpretation of randomized dual coordinate descent in the primal space. Also, in the special case of a strongly convex regularizer our generalized stochastic Frank–Wolfe method (as well as the randomized dual coordinate descent method) exhibits linear convergence. Furthermore, we present computational experiments that indicate that our method outperforms other stochastic Frank–Wolfe methods for a sufficiently small optimality tolerance, consistent with the theory developed herein. [ABSTRACT FROM AUTHOR]

The paper proposes the concept of inverse P-augmented matrix, and gives its structure; the inverse P-augmented matrix is obtained by improving the common augmented matrix; under certain condition, the inverse P-augmented matrix is reduced to the common augmented matrix. Inverse P-sets is the theoretical basis of proposing inverse P-augmented matrix, which is a new set model with dynamic feature. The paper gives the introduction of inverse-P set, the generation method of inverse P-augmented matrix and the relation between inverse P-augmented matrix and common augmented matrix, and gives the application of inverse P-augmented matrix in the dynamic findings of unknown information by virtue of these studies. [ABSTRACT FROM AUTHOR]

In order to reduce the loss of information of the majority class samples in the resampling process, combining the distribution of class samples and the characteristics of ensemble learning algorithm, in this paper, a two-level selective ensemble learning algorithm for imbalanced datasets is proposed. Firstly, the algorithm under-samples the majority class samples and constructs multiple training subsets. The training process will generate multiple base classifiers using AdaBoost algorithm, then select some base classifiers according to maximum correlation and minimum redundancy criteria, and form sub-classifiers according to weighted integration. Then, generate multiple sub-classifiers for multiple training subsets, and then, select some sub-classifiers according to maximum correlation and minimum redundancy criteria. Then, the weights of the selected sub-classifiers are calculated by F-means or G-means, and the ensemble classifier is obtained by weighted voting. Finally, the improved algorithm for imbalanced dataset is applied to the network anomaly detection. The experimental results on UCI datasets show that this method can improve the classification performance to a certain extent, especially for imbalanced datasets. Finally, the algorithm is applied to network anomaly detection for Internet of Things. From the simulation data of KDDCUP99 dataset, we can see that TLSE-ID algorithm has a small missing report rate and high precision. [ABSTRACT FROM AUTHOR]

Advances in data acquisition have generated an enormous amount of data that captures business, commercial, technological and scientific information. However, some occurrences are rare or unusual, irrespective of a large amount of data available. These rare occurrences in data mining are usually referred to as outliers or anomalies. All these rare occurrences are infrequent. Sometimes it varies from 0.01% to 10% depending on the type of application. In recent years, outlier detection has become important in many applications and has attracted considerable attention among the increasing number of data mining techniques. Focusing on this has resulted in several outlier detection algorithms, mostly based on distance or density. However, each method has its inherent weaknesses. Methods based on distance have problems with local density, and methods based on density have problems with low-density patterns. In this paper, we present a new outlier detection algorithm based on the relevant attribute analysis (ODRA) for local outlier detection in a high-dimensional dataset. There are two phases of the proposed algorithm. During the preliminary stage, we present a data reduction method that reduces the data set by pruning irrelevant attributes and data points. In the second phase, we propose an outlier detection method based on k-NN kernel density estimation. The experimental results on 15 UCI machine learning repository datasets show the supremacy and effectiveness of our proposed approach over state-of-the-art outlier detection methods. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) and machine learning (ML) tools play a significant role in the recent evolution of smart systems. AI solutions are pushing towards a significant shift in many fields such as healthcare, autonomous airplanes and vehicles, security, marketing customer profiling and other diverse areas. One of the main challenges hindering the AI potential is the demand for high-performance computation resources. Recently, hardware accelerators are developed in order to provide the needed computational power for the AI and ML tools. In the literature, hardware accelerators are built using FPGAs, GPUs and ASICs to accelerate computationally intensive tasks. These accelerators provide high-performance hardware while preserving the required accuracy. In this work, we present a systematic literature review that focuses on exploring the available hardware accelerators for the AI and ML tools. More than 169 different research papers published between the years 2009 and 2019 are studied and analysed. [ABSTRACT FROM AUTHOR]

Image registration is a fundamental task in image analysis in which the transform that moves the coordinate system of one image to another is calculated. Registration of multi-modal medical images has important implications for clinical diagnosis, treatment planning, and image-guided surgery as it provides the means of bringing together complimentary information obtained from different image modalities. However, since different image modalities have different properties due to their different acquisition methods, it remains a challenging task to find a fast and accurate match between multi-modal images. Furthermore, due to reasons such as ethical issues and need for human expert intervention, it is difficult to collect a large database of labelled multi-modal medical images. In addition, manual input is required to determine the fixed and moving images as input to registration algorithms. In this paper, we address these issues and introduce a registration framework that (1) creates synthetic data to augment existing datasets, (2) generates ground truth data to be used in the training and testing of algorithms, (3) registers (using a combination of deep learning and conventional machine learning methods) multi-modal images in an accurate and fast manner, and (4) automatically classifies the image modality so that the process of registration can be fully automated. We validate the performance of the proposed framework on CT and MRI images of the head obtained from a publicly available registration database. [ABSTRACT FROM AUTHOR]

This paper proposes a novel quantum learning algorithm based on Bernstein and Vazirani's quantum circuit to find the dependent variables of the 2-junta problem. Typically, for a given Boolean function f : {0, 1}n → {0, 1} that depends on only 2 out of n variables, the dependent variables are obtained by evaluating the function 4n times in the worst-case. However, the proposed quantum algorithm only requires O(log2n) function operations in the worst-case. Moreover, the algorithm requires an average of 5.3 function operations at the most when n ≥ 8. [ABSTRACT FROM AUTHOR]

Data clustering is one of the important techniques of data mining that is responsible for dividing N data objects into K clusters while minimizing the sum of intra-cluster distances and maximizing the sum of inter-cluster distances. Due to nonlinear objective function and complex search domain, optimization algorithms find difficulty during the search process. Recently, Harris hawks optimization (HHO) algorithm is proposed for solving global optimization problems. HHO has already proved its efficacy in solving a variety of complex problems. In this paper, a chaotic sequence-guided HHO (CHHO) has been proposed for data clustering. The performance of the proposed approach is compared against six state-of-the-art algorithms using 12 benchmark datasets of the UCI machine learning repository. Various comparative performance analysis and statistical tests have justified the effectiveness and competitiveness of the suggested approach. [ABSTRACT FROM AUTHOR]

Learning to rank (LtR) techniques leverage assessed samples of query-document relevance to learn effective ranking functions able to exploit the noisy signals hidden in the features used to represent queries and documents. In this paper we explore how to enhance the state-of-the-art LambdaMart LtR algorithm by integrating in the training process an explicit knowledge of the underlying user-interaction model and the possibility of targeting different objective functions that can effectively drive the algorithm towards promising areas of the search space. We enrich the iterative process followed by the learning algorithm in two ways: (1) by considering complex query-based user dynamics instead than simply discounting the gain by the rank position; (2) by designing a learning path across different loss functions that can capture different signals in the training data. Our extensive experiments, conducted on publicly available datasets, show that the proposed solution permits to improve various ranking quality measures by statistically significant margins. [ABSTRACT FROM AUTHOR]

Modeling the rate of penetration (ROP) plays a fundamental role in drilling optimization since the achievement of an optimum ROP can drastically reduce the overall cost of drilling activities. Evolved Extreme learning machine (ELM) with the evolutionary algorithms and multi-layer perceptron with Levenberg-Marquardt training algorithm (MLP-LMA) were proposed in this study to predict ROP. This paper focused mainly on two aspects. The first one was the investigation of the whale optimization algorithm (WOA) to optimize the weights and biases between input and hidden layers of ELM to enhance its prediction accuracy. The other was to adopt a prediction methodology that seeks to update the predictive model at each formation in order to reduce the dimension of input data and mitigate the effect of non real-time data such as the formation properties on the bit speed prediction. The prediction models were trained and tested using 3561 data points gathered from an Algerian field. The statistical and graphical evaluation criteria show that the ELM-WOA exhibited higher accuracy and generalization performance compared with the ELM-PSO and MLP-LMA. Furthermore, ELM-WOA was compared with two well-known ROP correlations in the literature, and the comparison results reveal that the proposed ELM-WOA model is superior to the pre-existing correlations. The findings of this study can help for the achievement of an optimum ROP and the reduction of the non-productive time. In addition, the outputs of this study can be used as an objective function during the real-time optimization of the drilling operation. [ABSTRACT FROM AUTHOR]

In this paper, a new metaheuristic algorithm called JAYA algorithm has been adapted for feature selection. Feature selection is a typical problem in machine learning and data mining domain concerned with determining the subset of high discriminative features from the irrelevant, noisy, redundant, and high-dimensional features. JAYA algorithm is initially proposed for continuous optimization. Due to the binary nature of the feature selection problem, the JAYA algorithm is adjusted using sinusoidal (i.e., S-shape) transfer function. Furthermore, the mutation operator controlled by adaptive mutation rate ( R m ) parameter is also utilized to control the diversity during the search. The proposed binary JAYA algorithm with adaptive mutation is called BJAM algorithm. The performance of BJAM algorithm is tested using 22 real-world benchmark datasets, which vary in terms of the number of features and the number of instances. Four measures are used for performance analysis: classification accuracy, number of features, fitness values, and computational times. Initially, a comparison between binary JAYA (BJA) algorithm and the proposed BJAM algorithm is conducted to show the effect of the mutation operator in the convergence behavior. After that, the results produced by the BJAM algorithm are compared against those yielded by ten state-of-the-art methods. Surprisingly, the proposed BJAM algorithm is able to excel other comparative methods in 7 out of 22 datasets in terms of classification accuracy. This can lead to the conclusion that the proposed BJAM algorithm is an efficient algorithm for the problems belonging to the feature selection domain and is pregnant with fruitful results. [ABSTRACT FROM AUTHOR]

Due to the huge amount of data being generating from different sources, the analyzing and extracting of useful information from these data becomes a very complex task. The difficulty of dealing with big data optimization problems comes from many factors such as the high number of features, and the existing of lost data. The feature selection process becomes an important step in many data mining and machine learning algorithms to reduce the dimensionality of the optimization problems and increase the performance of the classification or clustering algorithms. In this paper, a set of hybrid and efficient genetic algorithms are proposed to solve feature selection problem, when the handled data has a large feature size. The proposed algorithms use a new gene-weighted mechanism that can adaptively classify the features into strong relative features, weak or redundant features, and unstable features during the evolution of the algorithm. Based on this classification, the proposed algorithm gives the strong features high priority and the weak features less priority when generating new candidate solutions. In the same time, the proposed algorithm tries to more concentrate on unstable features that sometimes appear and sometimes disappear from the best solutions of the population. The performance of proposed algorithms is investigated by using different datasets and feature selection algorithms. The results show that our proposed algorithms can outperform the other feature selection algorithms and effectively enhance the classification performance over the tested datasets. [ABSTRACT FROM AUTHOR]

In extreme learning machine (ELM), a large number of hidden nodes are required due to the randomly generated hidden layer. To improve network compactness, the ELM with smoothed ℓ0 regularizer (ELM-SL0 for short) is studied in this paper. Firstly, the ℓ0 regularization penalty term is introduced into the conventional error function, such that the unimportant output weights are gradually forced to zeros. Secondly, the batch gradient method and the smoothed ℓ0 regularizer are combined for training and pruning ELM. Furthermore, both the weak convergence and strong convergence of ELM-SL0 are investigated. Compared with other existing ELMs, the proposed algorithm obtains better performance in terms of estimation accuracy and network sparsity. [ABSTRACT FROM AUTHOR]

In this paper, we present a new evolution-based algorithm that optimizes cell detection image processing workflows in a self-adaptive fashion. We use evolution strategies to optimize the parameters for all steps of the image processing pipeline and improve cell detection results. The algorithm reliably produces good cell detection results without the need for extensive domain knowledge. Our algorithm also needs no labeled data to produce good cell detection results compared to the state-of-the-art neural network approaches. Furthermore, the algorithm can easily be adapted to different applications by modifying the processing steps in the pipeline and has high scalability since it supports multithreading and computation on graphical processing units (GPUs). [ABSTRACT FROM AUTHOR]

The Goldreich–Levin algorithm was originally proposed for a cryptographic purpose and then applied to learning. The algorithm is to find some larger Walsh coefficients of an n variable Boolean function. Roughly speaking, it takes a p o l y (n , 1 ϵ log 1 δ) time to output the vectors w with Walsh coefficients S (w) ≥ ϵ with probability at least 1 - δ . However, in this paper, a quantum algorithm for this problem is given with query complexity O (log 1 δ ϵ 4) , which is independent of n. Furthermore, the quantum algorithm is generalized to apply for an n variable m output Boolean function F with query complexity O (2 m log 1 δ ϵ 4) . [ABSTRACT FROM AUTHOR]

One primary function in a cognitive radio network (CRN) is spectrum sensing. In an infrastructure-based CRN, instead of individual nodes independently sensing the presence of the incumbent signal and taking decisions thereon, a fusion center (FC) aggregates the sensing reports from the individual nodes and makes the final decision. Such collaborative spectrum sensing (CSS) is known to result in better sensing accuracy. On the other hand, CSS is vulnerable to Spectrum Sensing Data Falsification (SSDF) attack (a.k.a. Byzantine attack) wherein a node maliciously falsifies the sensing report prior to sending it to the FC, with the intention of disrupting the spectrum sensing process. This paper investigates the use of machine learning techniques, viz., SVM, Neural Network, Naive Bayes and Ensemble classifiers for detection of SSDF attacks in a CRN where the sensing reports are binary (i.e., either 1 or 0). The learning techniques are studied under two experimental scenarios: (a) when the training and test data are drawn from the same data-set, and (b) when separate data-sets are used for training and testing. Under the first scenario, of all the techniques, NN and Ensemble are the most robust showing consistently very good performance across varying presence of attackers in the system. Moreover performance comparison with an existing non-machine learning technique shows that the learning techniques are generally more robust than the existing algorithm under high presence of attackers. Under the second scenario, in a limited environment, Ensemble is the most robust method showing good overall performance. [ABSTRACT FROM AUTHOR]

In recent years, classifier ensemble techniques have drawn the attention of many researchers in the machine learning research community. The ultimate goal of these researches is to improve the accuracy of the ensemble compared to the individual classifiers. In this paper, a novel algorithm for building ensembles called dynamic programming-based ensemble design algorithm (DPED) is introduced and studied in detail. The underlying theory behind DPED is based on cooperative game theory in the first phase and applying a dynamic programming approach in the second phase. The main objective of DPED is to reduce the size of the ensemble while encouraging extra diversity in order to improve the accuracy. The performance of the DPED algorithm is compared empirically with the classical ensemble model and with a well-known algorithm called "the most diverse." The experiments were carried out with 13 datasets from UCI and three ensemble models. Each ensemble model is constructed from 15 different base classifiers. The experimental results demonstrate that DPED outperforms the classical ensembles on all datasets in terms of both accuracy and size of the ensemble. Regarding the comparison with the most diverse algorithm, the number of selected classifiers by DPED across all datasets and all domains is less than or equal to the number selected by the most diverse algorithm. Experiment on blog spam dataset, for instance, shows that DPED provides an accuracy of 96.47 compared to 93.87 obtained by the most diverse using 40% training size. Finally, the experimental results verify the reliability, stability, and effectiveness of the proposed DPED algorithm. [ABSTRACT FROM AUTHOR]

Clustering ensemble can overcome the instability of clustering and improve clustering performance. With the rapid development of clustering ensemble, we find that not all clustering solutions are effective in their final result. In this paper, we focus on selection strategy in selective clustering ensemble. We propose a multiple clustering and selecting approach (MCAS), which is based on different original clustering solutions. Furthermore, we present two combining strategies, direct combining and clustering combining, to combine the solutions selected by MCAS. These combining strategies combine results of MCAS and get a more refined subset of solutions, compared with traditional selective clustering ensemble algorithms and single clustering and selecting algorithms. Experimental results on UCI machine learning datasets show that the algorithm that uses multiple clustering and selecting algorithms with combining strategy performs well on most datasets and outperforms most selective clustering ensemble algorithms. [ABSTRACT FROM AUTHOR]

In this paper, an optimal weight semi-supervised learning machine for a single-hidden layer feedforward network (SLFN) with time delay is developed. Both input weights and output weights of the SLFN are globally optimized with manifold regularization. By feature mapping, input vectors can be placed at the prescribed positions in the feature space in the sense that the separability of all nonlinearly separable patterns can be maximized, unlabeled data can be leveraged to improve the classification accuracy when labeled data are scarce, and a high degree of recognition accuracy can be achieved with a small number of hidden nodes in the SLFN. Some simulation examples are presented to show the excellent performance of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Dimensionality reduction plays an important role in many machine learning and pattern recognition applications. Linear discriminant analysis (LDA) is the most popular supervised dimensionality reduction technique which searches for the projection matrix that makes the data points of different classes to be far from each other while requiring data points of the same class to be close to each other. In this paper, trace ratio LDA is combined with K-means clustering into a unified framework, in which K-means clustering is employed to generate class labels for unlabeled data and LDA is used to investigate low-dimensional representation of data. Therefore, by combining the subspace clustering with dimensionality reduction together, the optimal subspace can be obtained. Differing from other existing dimensionality reduction methods, our novel framework is suitable for different scenarios: supervised, semi-supervised, and unsupervised dimensionality reduction cases. Experimental results on benchmark datasets validate the effectiveness and superiority of our algorithm compared with other relevant techniques. [ABSTRACT FROM AUTHOR]