Showing papers in "Neural Processing Letters in 2021"
TL;DR: An investigational result shows that the proposed CCA-HBFNNC model can increases the sustainability level and minimizes the time complexity of urban development when contrasted with contemporary works.
Abstract: People give more importance concerning the overall quality of the modernized ecosystem. The pollution of air is one of the significant problems to be resolved as it restricted the ecological transformation of the modernized ecosystem. Therefore, it is fundamental to evaluate the implication of these ecological issues to enhance the urban ecosystem. This vital purpose of this research is to propose a canonical correlation analysis based hyper basis feedforward neural network classification (CCA-HBFNNC) model for evaluating sustainable urban environmental quality. The CCA-HBFNNC model initially acquires a large size of U.S. air pollution dataset as input. Then, a canonical correlative analysis based feature selection algorithm is applied in the CCA-HBFNNC model to select the key pollutant features, which bear fundamental implications to the modernize air pollution to maintain the level of urban sustainability. After the feature selection process, the CCA-HBFNNC model applies the HYPER BASIS FEEDFORWARD NEURAL NETWORK CLASSIFICATION (HBFNNC) algorithm in order to classify input air data based on chosen pollutants features. During the classification process, the HBFNNC algorithm used three critical layers namely hidden, output and input layers for efficiently categorizing each input data as higher or lower pollution level with higher accuracy. If the level of air pollution on the urban environment is higher, finally CCA-HBFNNC model significantly reduces the pollution level. In this way, the CCA-HBFNNC model attains improved urban sustainability levels when compared to sophisticated operation. An experimental evaluation of the CCA-HBFNNC model is determined in terms of CCA-HBFNNC model, time complexity and false-positive rate in consideration of the diversified number of air data retrieved from the big data sets. An investigational result shows that the proposed CCA-HBFNNC model can increases the sustainability level and minimizes the time complexity of urban development when contrasted with contemporary works.
165 citations
TL;DR: In this paper, the authors proposed a new model, which contains various layers in the aim to classify MRI brain tumor, and the proposed model is experimentally evaluated on three datasets.
Abstract: Brain tumor represents one of the most fatal cancers around the world. It is common cancer in adults and children. It has the lowest survival rate and various types depending on their location, texture, and shape. The wrong classification of the tumor brain will lead to bad consequences. Consequently, identifying the correct type and grade of tumor in the early stages has an important role to choose a precise treatment plan. Examining the magnetic resonance imaging (MRI) images of the patient’s brain represents an effective technique to distinguish brain tumors. Due to the big amounts of data and the various brain tumor types, the manual technique becomes time-consuming and can lead to human errors. Therefore, an automated computer assisted diagnosis (CAD) system is required. The recent evolution in image classification techniques has shown great progress especially the deep convolution neural networks (CNNs) which have succeeded in this area. In this regard, we exploited CNN for the problem of brain tumor classification. We suggested a new model, which contains various layers in the aim to classify MRI brain tumor. The proposed model is experimentally evaluated on three datasets. Experimental results affirm that the suggested approach provides a convincing performance compared to existing methods.
70 citations
TL;DR: A deep learning framework for skin cancer detection was applied to five state-of-art convolutional neural networks to create both a plain and a hierarchical (with 2 levels) classifiers that are capable to distinguish between seven types of moles.
Abstract: Skin diseases have become a challenge in medical diagnosis due to visual similarities. Although melanoma is the best-known type of skin cancer, there are other pathologies that are the cause of many death in recent years. The lack of large datasets is one of the main difficulties to develop a reliable automatic classification system. This paper presents a deep learning framework for skin cancer detection. Transfer learning was applied to five state-of-art convolutional neural networks to create both a plain and a hierarchical (with 2 levels) classifiers that are capable to distinguish between seven types of moles. The HAM10000 dataset, a large collection of dermatoscopic images, were used for experiments, with the help of data augmentation techniques to improve performance. Results demonstrate that the DenseNet201 network is suitable for this task, achieving high classification accuracies and F-measures with lower false negatives. The plain model performed better than the 2-levels model, although the first level, i.e. a binary classification, between nevi and non-nevi yielded the best outcomes.
55 citations
TL;DR: This work provides a foundation in spiking signal encoding and gives an overview over different application-oriented implementations which utilise the schemes.
Abstract: Biologically inspired spiking neural networks are increasingly popular in the field of artificial intelligence due to their ability to solve complex problems while being power efficient. They do so by leveraging the timing of discrete spikes as main information carrier. Though, industrial applications are still lacking, partially because the question of how to encode incoming data into discrete spike events cannot be uniformly answered. In this paper, we summarise the signal encoding schemes presented in the literature and propose a uniform nomenclature to prevent the vague usage of ambiguous definitions. Therefore we survey both, the theoretical foundations as well as applications of the encoding schemes. This work provides a foundation in spiking signal encoding and gives an overview over different application-oriented implementations which utilise the schemes.
53 citations
TL;DR: The results of the experiment prove that the proposed algorithm of feature selection that is wrapper-based is capable of achieving better levels of performance compared to existing methods like minimum redundancy maximum relevance, and correlation-based feature selection.
Abstract: The symptoms of cancer normally appear only in the advanced stages, so it is very hard to detect resulting in a high mortality rate among the other types of cancers. Thus, there is a need for early prediction of lung cancer for the purpose of diagnosing and this can result in better chances of it being able to be treated successfully. Histopathology images of lung scan can be used for classification of lung cancer using image processing methods. The features from lung images are extracted and employed in the system for prediction. Grey level co-occurrence matrix along with the methods of Gabor filter feature extraction are employed in this investigation. Another important step in enhancing the classification is feature selection that tends to provide significant features that helps differentiating between various classes in an accurate and efficient manner. Thus, optimal feature subsets can significantly improve the performance of the classifiers. In this work, a novel algorithm of feature selection that is wrapper-based is proposed by employing the modified stochastic diffusion search (SDS) algorithm. The SDS, will benefit from the direct communication of agents in order to identify optimal feature subsets. The neural network, Naive Bayes and the decision tree have been used for classification. The results of the experiment prove that the proposed method is capable of achieving better levels of performance compared to existing methods like minimum redundancy maximum relevance, and correlation-based feature selection.
50 citations
TL;DR: A region centric minutiae propagation measure (RCMPM) based approach for forged finger print detection has been improved and reduces the false classification ratio.
Abstract: The problem of forgery detection has been well studied and the forged finger prints produces highly impacting results in the biometric based security systems. There are many algorithms discussed earlier to detect forged finger prints. However, they suffer to achieve higher performance in terms of security. In this paper, a region centric minutiae propagation measure (RCMPM) based approach. First, the finger print image is read and removes the noisy points by applying the multi level Gabor filters. The Gabor filter has been applied in multiple level which helps to remove the noise from finger print image. The enhanced image is converted into number of integral image. The integral images are generated by splitting the image into number of tiny images according to the size of window. From the integral image produced, the island, dot, enclosure, bifurcation features are extracted. Extracted features are framed as feature vector and used to estimate the RCMPM measure. Based on the RCMPM measure, the presence of forged finger print has been identified and the same has been used to identify the region which has been modified. The accuracy of forged print detection has been improved and reduces the false classification ratio.
50 citations
TL;DR: In this paper, a Nonlinear Autoregressive Neural Network Time Series (NAR-NNTS) model is proposed for predicting confirmed, recovered and death cases of COVID-19 outbreak.
Abstract: The recent COVID-19 outbreak has severely affected people around the world. There is a need of an efficient decision making tool to improve awareness about the spread of COVID-19 infections among the common public. An accurate and reliable neural network based tool for predicting confirmed, recovered and death cases of COVID-19 can be very helpful to the health consultants for taking appropriate actions to control the outbreak. This paper proposes a novel Nonlinear Autoregressive (NAR) Neural Network Time Series (NAR-NNTS) model for forecasting COVID-19 cases. This NAR-NNTS model is trained with Scaled Conjugate Gradient (SCG), Levenberg Marquardt (LM) and Bayesian Regularization (BR) training algorithms. The performance of the proposed model has been compared by using Root Mean Square Error (RMSE), Mean Square Error (MSE) and correlation co-efficient i.e. R-value. The results show that NAR-NNTS model trained with LM training algorithm performs better than other models for COVID-19 epidemiological data prediction.
49 citations
TL;DR: In this article, the robust passivity and stability analysis of uncertain complex-valued impulsive neural network (UCVINN) models with time-varying delays are investigated. But the authors consider the uncertainty of norm-bounded parameters to achieve more realistic system behaviors.
Abstract: In this article, we investigate the robust passivity and stability analysis of uncertain complex-valued impulsive neural network (UCVINN) models with time-varying delays. Many practical systems are subject to uncertainty in the real-world environments. As a result, we consider the uncertainty of norm-bounded parameters to achieve more realistic system behaviors. By using appropriate Lyapunov–Krasovskii functionals and integral inequalities, sufficient conditions for the robust passivity and global asymptotic stability of UCVINNs are derived by separating complex-valued neural networks into real and imaginary parts. The criteria are given in terms of linear matrix inequalities (LMIs) that can be checked by the MATLAB LMI toolbox. Finally, numerical simulations are presented to illustrate the merits of the obtained results.
48 citations
TL;DR: With the proposed approach an individual can predict the insulin resistance and hence prospective chances of diabetes might be tracked daily using non-clinical approaches while the same is not practically possible with clinical processes daily.
Abstract: Identification and quantification of insulin resistance require specific blood test which is complex, time-consuming, and much more invasive, making it difficult to track the changes daily. With the advancement in machine learning approaches, identification of insulin resistance can be carried out without clinical processes. In this work, insulin resistance is identified for individuals with triglycerides and HDL-c ratio using non-invasive techniques employing machine learning approaches. Eighteen parameters are used for identification purposes like age, sex, waist size, height, etc., and combinations of these parameters. Experiments are conducted over the CALERIE dataset. Each output of the attribute selection system is modeled over distinct calculations like logistic regression, CARTs, SVM, LDA, KNN, extra trees classifier. The proposed work is validated with a stratified cross-validation test. Results show that KNN and CatBoost show the best results with an accuracy of 74% and 73% respectively and 1% variance compared to 66% with Bernardini et al. and Stawiski et al. and 83% with Farran et al. With the proposed approach an individual can predict the insulin resistance and hence prospective chances of diabetes might be tracked daily using non-clinical approaches. While the same is not practically possible with clinical processes daily.
46 citations
TL;DR: A combination of Artificial Neural Network and Fuzzy K-means algorithm has been presented to segment the tumor locale and the overall accuracy has been improved by 8% when compared with K-Nearest Neighbor methodology.
Abstract: The primary objective of this paper is to develop a methodology for brain tumor segmentation. Nowadays, brain tumor recognition and fragmentation is one among the pivotal procedure in surgical and medication planning arrangements. It is difficult to segment the tumor area from MRI images due to inaccessibility of edge and appropriately visible boundaries. In this paper, a combination of Artificial Neural Network and Fuzzy K-means algorithm has been presented to segment the tumor locale. It contains four phases, (1) Noise evacuation (2) Attribute extraction and selection (3) Classification and (4) Segmentation. Initially, the procured image is denoised utilizing wiener filter, and then the significant GLCM attributes are extricated from the images. Then Deep Learning based classification has been performed to classify the abnormal images from the normal images. Finally, it is processed through the Fuzzy K-Means algorithm to segment the tumor region separately. This proposed segmentation approach has been verified on BRATS dataset and produces the accuracy of 94%, sensitivity of 98% specificity of 99%, Jaccard index of 96%. The overall accuracy of this proposed technique has been improved by 8% when compared with K-Nearest Neighbor methodology.
29 citations
TL;DR: In this article, the authors proposed a framework named PP-SPA for privacy-preserved human activity recognition (HAR) and real-time activity functioning support using the smartphone-based virtual personal assistant.
Abstract: Mobile ubiquitous computing has not only enriched human comfort but also has a deep impact on changing standards of human daily life. Modern inventions can be even more automated by using the Internet of Things (IoT) and Artificial Intelligence (AI). Mobile devices, body area networks, and embedded computing systems allow healthcare providers to continuously assess and monitor their patients but also bring privacy concerns. This paper proposes a smartphone-based end-to-end novel framework named PP-SPA for privacy-preserved Human Activity Recognition (HAR) and real-time activity functioning support using the smartphone-based virtual personal assistant. PP-SPA helps to improve the routine life functioning of the Cognitive Impaired individuals. PP-SPA uses a highly accurate machine learning model that takes input from smartphone sensors (i.e., accelerometer, gyroscope, magnetometer, and GPS) for accurate HAR and uses a digital diary to recommend real-time support for improvement in individual’s health. PP-SPA yields a proficient accuracy of 90% with the Hoeffding Tree and Logistic Regression algorithm which endeavors reasonable models in terms of uncertainty.
TL;DR: The improved algorithm provides a more superior compared with the basic RRT algorithm and Bg-RRT algorithm, and has shorter computational time and shorter path than the traditional R RT algorithm.
Abstract: A 1–0 Bg-RRT algorithm is proposed to reduce computational time and complexity, even in complex environments. Different from Rapidly-exploring Random Tree (RRT) and Bias-goal Rapidly-exploring Random Tree (Bg-RRT), using 1–0 Bg-RRT with 1 and 0 change probability biased to the target to construct the tree is faster and can jump out of the local minimum in time. Although unknown space path planning problem based on RRT is difficult to obtain satisfactory performance, but the improved algorithm provides a more superior compared with the basic RRT algorithm and Bg-RRT algorithm. The simulation results show that the 1–0 Bg-RRT algorithm has shorter computational time and shorter path than the traditional RRT algorithm.
TL;DR: By using zeroing neural dynamics method, a continuous-time model is proposed for solving the time-varying problem of QRD in real-time by using time derivative information from a known real or complex matrix.
Abstract: QR decomposition (QRD) is of fundamental importance for matrix factorization in both real and complex cases. In this paper, by using zeroing neural dynamics method, a continuous-time model is proposed for solving the time-varying problem of QRD in real-time. The proposed dynamics use time derivative information from a known real or complex matrix. Furthermore, its theoretical analysis is provided to substantiate the convergence and effectiveness of solving the time-varying QRD problem. In addition, numerical experiments in four different-dimensional time-varying matrices show that the proposed model is effective for solving the time-varying QRD problem both in the case of a real or a complex matrix as input.
TL;DR: This article proposed a monolingual BERT for the Persian language (ParsBERT), which shows its state-of-the-art performance compared to other architectures and multilingual models.
Abstract: The surge of pre-trained language models has begun a new era in the field of Natural Language Processing (NLP) by allowing us to build powerful language models. Among these models, Transformer-based models such as BERT have become increasingly popular due to their state-of-the-art performance. However, these models are usually focused on English, leaving other languages to multilingual models with limited resources. This paper proposes a monolingual BERT for the Persian language (ParsBERT), which shows its state-of-the-art performance compared to other architectures and multilingual models. Also, since the amount of data available for NLP tasks in Persian is very restricted, a massive dataset for different NLP tasks as well as pre-training the model is composed. ParsBERT obtains higher scores in all datasets, including existing ones and gathered ones, and improves the state-of-the-art performance by outperforming both multilingual BERT and other prior works in Sentiment Analysis, Text Classification, and Named Entity Recognition tasks.
TL;DR: This work proposes a spectral new clustering method to consider the feature selection with the L_{2,1}$$ -norm regularization as well as simultaneously learns orthogonal representations for each sample to preserve the local structures of data points.
Abstract: Redundant features and outliers (noise) included in the data points for a machine learning clustering model heavily influences the discovery of more distinguished features for clustering. To solve this issue, we propose a spectral new clustering method to consider the feature selection with the $$L_{2,1}$$
-norm regularization as well as simultaneously learns orthogonal representations for each sample to preserve the local structures of data points. Our model also solves the issue of out-of-sample, where the training process does not output an explicit model to predict unseen data points, along with providing an efficient optimization method for the proposed objective function. Experimental results showed that our method on twelve data sets achieves the best performance compared with other similar models.
TL;DR: This article focuses on diagnosing a diabetic patient through data mining techniques and indicates that the proposed method with an accuracy of 91.65% outperformed the results of the earlier methods examined in this article.
Abstract: Artificial intelligence is a future and valuable tool for early disease recognition and support in patient condition monitoring. It can increase the reliability of the cure and decision making by developing useful systems and algorithms. Healthcare workers, especially nurses and physicians, are overworked due to a massive and unexpected increase in the number of patients during the coronavirus pandemic. In such situations, artificial intelligence techniques could be used to diagnose a patient with life-threatening illnesses. In particular, diseases that increase the risk of hospitalization and death in coronavirus patients, such as high blood pressure, heart disease and diabetes, should be diagnosed at an early stage. This article focuses on diagnosing a diabetic patient through data mining techniques. If we are able to diagnose diabetes in the early stages of the disease, we can force patients to stay home and care for their health, so the risk of being infected with the coronavirus would be reduced. The proposed method has three steps: preprocessing, feature selection and classification. Several combinations of Harmony search algorithm, genetic algorithm, and particle swarm optimization algorithm are examined with K-means for feature selection. The combinations have not examined before for diabetes diagnosis applications. K-nearest neighbor is used for classification of the diabetes dataset. Sensitivity, specificity, and accuracy have been measured to evaluate the results. The results achieved indicate that the proposed method with an accuracy of 91.65% outperformed the results of the earlier methods examined in this article.
TL;DR: The experimental results show that the evaluation criteria in ETOR have improved compared to other secure routing algorithms, and the ETOR algorithm provides more optimal routes for the effective transmission than its untrusted version, i.e. the energy-aware opportunity based routing.
Abstract: Rapid developments in processors and radio technology have led to the emergence of small sensor nodes capable of communicating in wireless sensor networks (WSNs). Nodes in WSN transmit data using multi-hop routing and based on cooperation with each other. This collaboration has made these types of networks vulnerable to many attacks. In order to determine the reliability of nodes in separating malicious nodes from other nodes, an intelligent trust management scheme must be used. In recent years, trust-based routing protocols and opportunistic routing have become important tools to increase WSN security and performance. In this paper, an energy-aware trust and opportunity-based routing (ETOR) algorithm is proposed with respect to a novel hybrid fitness function. This algorithm has two main steps: one is to select secure nodes based on tolerance constant and the other is to select opportunistic nodes from secure nodes to perform routing. ETOR uses the multipath routes technique with an intra-cluster and inter-cluster multi-hop communication mechanism. In addition, the optimal and secure route is selected based on a novel hybrid fitness function with the parameters of energy, trust, QoS, connectivity, distance, hop-count and network traffic. The simulation was performed by MATLAB based on evaluation criteria such as throughput, delay, detection rate, NRL, distance, energy, packet delivery ratio and network lifetime in the presence of DoS attack. The experimental results show that the evaluation criteria in ETOR have improved compared to other secure routing algorithms. In addition, the ETOR algorithm provides more optimal routes for the effective transmission than its untrusted version, i.e. the energy-aware opportunity based routing.
TL;DR: A new neural network model is proposed to improve the effectiveness of the NER by using a pre-trained XLNet, bi-directional long-short term memory (Bi-LSTM) and conditional random field (CRF) and the superiority of XLNet in NER tasks is demonstrated.
Abstract: Named entity recognition (NER) is the basis for many natural language processing (NLP) tasks such as information extraction and question answering. The accuracy of the NER directly affects the results of downstream tasks. Most of the relevant methods are implemented using neural networks, however, the word vectors obtained from a small data set cannot describe unusual, previously-unseen entities accurately and the results are not sufficiently accurate. Recently, the use of XLNet as a new pre-trained model has yielded satisfactory results in many NLP tasks, integration of XLNet embeddings in existent NLP tasks is not straightforward. In this paper, a new neural network model is proposed to improve the effectiveness of the NER by using a pre-trained XLNet, bi-directional long-short term memory (Bi-LSTM) and conditional random field (CRF). Pre-trained XLNet model is used to extract sentence features, then the classic NER neural network model is combined with the obtained features. In addition, the superiority of XLNet in NER tasks is demonstrated. We evaluate our model on the CoNLL-2003 English dataset and WNUT-2017 and show that the XLNet-BiLSTM-CRF obtains state-of-the-art results.
TL;DR: A new model of controlling DOA with no need for the use of such an index is proposed applying a feedforward neural network and an adaptive neuro-fuzzy inference model resulting in optimal drug dose and stable anesthesia depth.
Abstract: This study aims to estimate the depth of anesthesia (DOA) at a safe and appropriate level taking into account the patient characteristics during the induction phase. Bi-spectral Index signal (BIS) as a common approach of controlling DOA generates noise and delays in the initial phase of induction. This may lead to useless information in the process of controlling. Moreover, using the BIS index entails a time-consuming process, high equipping costs, and a lack of accessibility to device accessories. To overcome these problems, we propose a new model of controlling DOA with no need for the use of such an index. Hence, an estimation strategy for DOA is developed applying a feedforward neural network and an adaptive neuro-fuzzy inference estimation model. This model estimates the dose of intravenous anesthetic drugs concerning the patients’ needs resulting in optimal drug dose and stable anesthesia depth. The proposed estimations are tested by sensitivity analysis being compared with real data obtained from the classical model (PK-PD) revised approach and BIS approach on 13 patients undergoing surgery. The results show an accuracy of 0.999, indicative of a high-validated model. Compared to BIS, our proposed model not only controls DOA accurately but also achieves outcomes in practice successfully. Some practical implications for future research and clinical practice are also suggested.
TL;DR: A semantic fake news detection method built around relational features like sentiment, entities or facts extracted directly from text is discussed, showing that adding semantic features improves accuracy significantly.
Abstract: The nuances of languages, as well as the varying degrees of truth observed in news items, make fake news detection a difficult problem to solve. A news item is never launched without a purpose, therefore in order to understand its motivation it is best to analyze the relations between the speaker and its subject, as well as different credibility metrics. Inferring details about the various actors involved in a news item is a problem that requires a hybrid approach that mixes machine learning, semantics and natural language processing. This article discusses a semantic fake news detection method built around relational features like sentiment, entities or facts extracted directly from text. Our experiments are focused on short texts with different degrees of truth and show that adding semantic features improves accuracy significantly.
TL;DR: Wang et al. as mentioned in this paper presented a secure business method with key distribution cryptographic techniques for payments via mobile for the e health care application The proposed method takes benefit of the advantages of payment using android and a refined key distribution cryptosystem to concurrently provide ehealth care business safety and attain payment efficiency in day to day life.
Abstract: Through the growing attractiveness of the financial world, the e health care application has developed quicker than the previous period, such that mobile payment adore extraordinary fame and are occupying an ever-growing business This is especially factual of mobile expenditures, which are interesting growing attention Though, the occasion of various old-style financial misfortunes has visible the tasks in-built in virtual verification technology that is based on old-style methods of understanding the strong and steady growth of payment via mobile In count, this technology guarantees data user account safety and confidentiality In this paper, we propose a Secure Authentication Protocol (SAP) payment via mobile To assurance trustworthy service, we use cryptographic techniques for attaining common authentication among the server as well as client, which can attack forged servers and fake workstations Related to the styles presently used, the proposed method supports the safety of data user account data as well as distinct privacy during the payment business progression via mobile To concurrently attain safety sturdiness and continue the practice suitability of payments via mobile within unconfident public communique systems is a critical matter for smart mobile device producers in addition to mobile data users In this paper, we present a secure business method with key distribution cryptographic techniques for payments via mobile for the e health care application The proposed method takes benefit of the advantages of payment using android and a refined key distribution cryptosystem to concurrently provide e health care business safety and attain payment efficiency in day to day life With a properly defined challenger prototype and safety analysis, the proposed method is confirmed to be mutually correct and safe by using the key distribution method It delivers robust business healthiness and communique safety to mobile data users for the duration of virtual payment dealings Alternatively, the performance analysis demonstrations that our proposed transaction method has a low computation cost when compared to the previous papers
TL;DR: This work designed a novel spiking deep convolutional neural network-based pipeline to classify AD using MRI scans and outperformed almost all the comparable methods due to the robust discriminative capability of the SNN in extracting relevant AD features for the AD classification task.
Abstract: Diagnosing Alzheimer’s Disease (AD) in older people using magnetic resonance imaging (MRI) is quite hard since it requires the extraction of highly discriminative feature representation from similar brain patterns and pixel intensities. However, deep learning techniques possess the capability of extracting relevant representations from data. In this work, we designed a novel spiking deep convolutional neural network-based pipeline to classify AD using MRI scans. We considered three MRI scan groups (patients with AD dementia, Mild Cognitive Impairment (MCI), and healthy controls (NC)). We developed a three-binary classification task (AD vs. NC, AD vs. MCI, and NC vs. MCI) for the AD classification tasks. Specifically, an unsupervised convolutional Spiking Neural Networks (SNN) is pre-trained on the MRI scans. Finally, a supervised deep Convolution Neural Network (CNN) is trained on the output of the SNN for the classification tasks. Experiments are performed using the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset, and promising results are obtained for the AD classification tasks. We present our proposed model results for both the unsupervised spike pre-training technique and the case where the pre-training technique was not considered, thus serving as a baseline. The accuracy of the proposed model with spike pre-training techniques for the three-binary classification are 90.15%, 87.30%, and 83.90%, respectively, and the accuracy of the model without the spike are 86.90%, 83.25%, and 76.70%, respectively, with a noticeable increase in accuracy and thus, reveals the effectiveness of the proposed method. We also evaluated the robustness of our proposed approach by running experiment on six baseline methods using our preprocessed MRI scans. Our model outperformed almost all the comparable methods due to the robust discriminative capability of the SNN in extracting relevant AD features for the AD classification task.
TL;DR: Wang et al. as mentioned in this paper proposed a density-weighted least squares twin support vector machine (DWLSTSVM) for binary class imbalance learning, where the optimization problem is solved by simply considering the equality constraints and by considering the 2-norm of slack variables.
Abstract: Usually the real-world (RW) datasets are imbalanced in nature, i.e., there is a significant difference between the number of negative and positive class samples in the datasets. Because of this, most of the conventional classifiers do not perform well on RW data classification problems. To handle the class imbalanced problems in RW datasets, this paper presents a novel density-weighted twin support vector machine (DWTWSVM) for binary class imbalance learning (CIL). Further, to boost the computational speed of DWTWSVM, a density-weighted least squares twin support vector machine (DWLSTSVM) is also proposed for solving the CIL problem, where, the optimization problem is solved by simply considering the equality constraints and by considering the 2-norm of slack variables. The key ideas behind the models are that during the model training phase, the training data points are given weights based on their importance, i.e., the majority class samples are given more importance compared to the minority class samples. Simulations are carried on a synthetic imbalanced and some real-world imbalanced datasets. The model performance in terms of F1-score, G-mean, recall and precision of the proposed DWTWSVM and DWLSTSVM are compared with support vector machine (SVM), twin SVM (TWSVM), least squares TWSVM (LSTWSVM), fuzzy TWSVM (FTWSVM), improved fuzzy least squares TWSVM (IFLSTWSVM) and density-weighted SVM for binary CIL. Finally, a statistical study is carried out based on F1-score and G-mean on RW datasets to verify the efficacy and usability of the suggested models.
TL;DR: A frame was formed with raw samples of the same activity which were collected consecutively from the accelerometer sensor to capture the pattern inherent in the activity and due to preserving the continuous structure of the movement.
Abstract: With the spread of wearable sensors, the solutions to the task of activity recognition by using the data obtained from the sensors have become widespread. Recognition of activities owing to wearable sensors such as accelerometers, gyroscopes, and magnetometers, etc. has been studied in recent years. Although there are several applications in the literature, differently in this study, deep learning algorithms such as Convolutional Neural Networks, Convolutional LSTM, and 3D Convolutional Neural Networks fed by Convolutional LSTM have been used in human activity recognition task by feeding with data obtained from accelerometer sensor. For this purpose, a frame was formed with raw samples of the same activity which were collected consecutively from the accelerometer sensor. Thus, it is aimed to capture the pattern inherent in the activity and due to preserving the continuous structure of the movement.
TL;DR: A deep learning CNN model is developed using multi-scale input with multiple stages of convolutional layers and a sigmoid activation function is used to classify pixels into forged or non-forged using the final feature map.
Abstract: Images are an important source of information and copy-move forgery (CMF) is one of the vicious forgery attacks. Its objective is to conceal sensitive information from the image. Hence, authentication of an image from human eyes become arduous. Reported techniques in literature for detection of CMF are suffering from the limitations of geometric transformations of forged region and computation cost. In this paper, a deep learning CNN model is developed using multi-scale input with multiple stages of convolutional layers. These layers are divided into two blocks i.e. encode and decoder. In encoder block, extracted feature maps from convolutional layers of multiple stages are combined and down sampled. Similarly, in decoder block extracted feature maps are combined and up sampled. A sigmoid activation function is used to classify pixels into forged or non-forged using the final feature map. To validate the model two different publicly available datasets are used. The performance of the proposed model is compared with state-of-the-art methods which show that the presented data-driven approach is better.
TL;DR: By utilizing the stability theory of impulsive systems, algebraic graph theory, sufficient conditions are derived to guarantee the leader-following consensus of the multi-agent system.
Abstract: In this manuscript, the stability and consensus problem of the leader-following multi-agent system is studied. The state information of the leader is only available to the dynamics of all the followers, while the communication among the agents occurs at sampling instant. The innovative part of this paper is to reach the consensus and attain the stability for prescribed MASs with interval time varying delay using impulsive control in the presence of leader. The interaction between the agent is illustrated by an undirected graph. A class of distributed impulsive protocol relying on sampling information is suggested to reach the leader following consensus. The consensus is critically relying on the sampling period, control gains. By performing the similar procedures, the consensus problem of multi-agent system is converted to stability problem of the error system. By utilizing the stability theory of impulsive systems, algebraic graph theory, sufficient conditions are derived to guarantee the leader-following consensus of the multi-agent system. Terminally, a numerical example is given to show an efficacy of this presented approach and the sureness of theoretical analysis.
TL;DR: The finite time synchronization problem of fractional order quaternion valued neural networks with time delay is investigated through using Lyapunov direct method and the setting time is estimated, which is influenced by the order of fractionAL derivative and control parameters.
Abstract: In this article, the finite time (FT) synchronization problem of fractional order quaternion valued neural networks with time delay is investigated. Without separating the quaternion valued system into two complex valued or four real valued systems, the FT synchronization conditions are derived through using Lyapunov direct method. Furthermore, the setting time is estimated, which is influenced by the order of fractional derivative and control parameters. Finally, numerical simulations are shown to verify the effectiveness of the proposed methods.
TL;DR: Experiments on four Amazon (5-core) datasets demonstrated that the proposed CSSAE model persistently outperforms state-of-the-art top-N recommendation methods on various effectiveness metrics.
Abstract: Context-aware recommender systems (CARS) are a vital module of many corporate, especially within the online commerce domain, where consumers are provided with recommendations about products potentially relevant for them. A traditional CARS, which utilizes deep learning models considers that user’s preferences can be predicted by ratings, reviews, demographics, etc. However, the feedback given by the users is often conflicting when comparing the rating score and the sentiment behind the reviews. Therefore, a model that utilizes either ratings or reviews for predicting items for top-N recommendation may generate unsatisfactory recommendations in many cases. In order to address this problem, this paper proposes an effective context-specific sentiment based stacked autoencoder (CSSAE) to learn the concrete preference of the user by merging the rating and reviews for a context-specific item into a stacked autoencoder. Hence, the user's preferences are consistently predicted to enhance the Top-N recommendation quality, by adapting the recommended list to the exact context where an active user is operating. Experiments on four Amazon (5-core) datasets demonstrated that the proposed CSSAE model persistently outperforms state-of-the-art top-N recommendation methods on various effectiveness metrics.
TL;DR: In this article, two efficient approaches of twin support vector machines (TWSVM) were proposed for binary classification, namely LTWSVM and GTWSVM, which reformulate the TWSVM formulation by introducing $$L_1$$¯¯ and $$L\infty $$ norms in the objective functions, and convert into linear programming problems.
Abstract: In this paper, we propose two efficient approaches of twin support vector machines (TWSVM). The first approach is to reformulate the TWSVM formulation by introducing $$L_1$$
and $$L_\infty $$
norms in the objective functions, and convert into linear programming problems termed as LTWSVM for binary classification. The second approach is to solve the primal TWSVM, and convert into completely unconstrained minimization problem. Since the objective function is convex, piecewise quadratic but not twice differentiable, we present an efficient algorithm using the generalized Newton’s method termed as GTWSVM. Computational comparisons of the proposed LTWSVM and GTWSVM on synthetic and several real-world benchmark datasets exhibits significantly better performance with remarkably less computational time in comparison to relevant baseline methods.
TL;DR: Wang et al. as discussed by the authors proposed a graph learning method to generate high-quality topological graph, which is more suitable for GCNs model classification, using the correlation between the data to generate a data similarity matrix, and apply Laplacian rank constraint to similarity matrix.
Abstract: Graph convolutional networks (GCNs), as an extension of classic convolutional neural networks (CNNs) in graph processing, have achieved good results in completing semi-supervised learning tasks. Traditional GCNs usually use fixed graph to complete various semi-supervised classification tasks, such as chemical molecules and social networks. Graph is an important basis for the classification of GCNs model, and its quality has a large impact on the performance of the model. For low-quality input graph, the classification results of the GCNs model are often not ideal. In order to improve the classification effect of GCNs model, we propose a graph learning method to generate high-quality topological graph, which is more suitable for GCNs model classification. We use the correlation between the data to generate a data similarity matrix, and apply Laplacian rank constraint to similarity matrix, so that the number of connected components of the topological graph is consistent with the number of categories of the original data. Experimental results on 10 real datasets show that our method is better than the comparison method in classification effect.