Neural Network Classification Method Research Articles

Task-State EEG Signal Classification for Spatial Cognitive Evaluation Based on Multiscale High-Density Convolutional Neural Network.

In this study, a multi-scale high-density convolutional neural network (MHCNN) classification method for spatial cognitive ability assessment was proposed, aiming at achieving the binary classification of task-state EEG signals before and after spatial cognitive training. Besides, the multi-dimensional conditional mutual information method was used to extract the frequency band features of the EEG data. And the coupling features under the combination of multi-frequency bands were transformed into multi-spectral images. At the same time, the idea of Densenet was introduced to improve the multi-scale convolutional neural network. Firstly, according to the discreteness of multispectral EEG image features, two-scale convolution kernels were used to calculate and learn useful channel and frequency band feature information in multispectral image data. Secondly, to enhance feature propagation and reduce the number of parameters, the dense network was connected after the multi-scale convolutional network, and the learning rate change function of the stochastic gradient descent algorithm was optimized to objectively evaluate the training effect. The experimental results showed that compared with the classical convolution neural network (CNN) and multi-scale convolution neural network, the proposed MHCNN had better classification performance in the six frequency band combinations with the highest accuracy of 98%: Theta-Alpha2-Gamma, Alpha2-Beta2-Gamma, Beta1-Beta2-Gamma, Theta-Beta2-Gamma, Theta- Alpha1-Gamma, and Alpha1-Alpha2-Gamma. By comparing the classification results of six frequency band combinations, it was found that the combination of the Theta-Beta2-Gamma band had the best classification effect. The MHCNN classification method proposed in this research could be used as an effective biological indicator of spatial cognitive training effect and could be extended to other brain function evaluations.

Second-order information bottleneck based spiking neural networks for sEMG recognition

The pattern recognition of surface electromyography (sEMG) signal is an important application in the realization of human-machine interface. However, due to the disturbance of human body, sensors and environment, sEMG signal usually contains lots of noise, which brings great challenges to the high-accuracy sEMG pattern recognition. In addition, embedded human wearable devices are becoming more and more popular nowadays. How to realize the sEMG recognition method with low power consumption and high noise immunity has also become a difficult and very meaningful research topic. In this paper, a spiking neural network (SNN) classification method based on second-order information bottleneck training is proposed. Firstly, the training loss function for classification neural networks is constructed based on the proposed second-order information bottleneck. The method is used to train the conventional continuous-valued neural network and convert it into an SNN model with equivalent structure and connection weights. Then, the converted SNN is used to classify the sEMG signal patterns. Through a series of theoretical analysis and experimental results, it is proved that this method has significant advantages in terms of generalization of network determination and computational efficiency. The experimental code can be accessed fromhttps://github.com/anvien/2OIB-for-sEMG-Recognition.

End-to-End Neural Network for Feature Extraction and Cancer Diagnosis of In Vivo Fluorescence Lifetime Images of Oral Lesions.

In contrast to previous studies that focused on classical machine learning algorithms and hand-crafted features, we present an end-to-end neural network classification method able to accommodate lesion heterogeneity for improved oral cancer diagnosis using multispectral autofluorescence lifetime imaging (maFLIM) endoscopy. Our method uses an autoencoder framework jointly trained with a classifier designed to handle overfitting problems with reduced databases, which is often the case in healthcare applications. The autoencoder guides the feature extraction process through the reconstruction loss and enables the potential use of unsupervised data for domain adaptation and improved generalization. The classifier ensures the features extracted are task-specific, providing discriminative information for the classification task. The data-driven feature extraction method automatically generates task-specific features directly from fluorescence decays, eliminating the need for iterative signal reconstruction. We validate our proposed neural network method against support vector machine (SVM) baselines, with our method showing a 6.5%-8.3% increase in sensitivity. Our results show that neural networks that implement data-driven feature extraction provide superior results and enable the capacity needed to target specific issues, such as inter-patient variability and the heterogeneity of oral lesions.Clinical relevance- We improve standard classification algorithms for in vivo diagnosis of oral cancer lesions from maFLIm for clinical use in cancer screening, reducing unnecessary biopsies and facilitating early detection of oral cancer.

Population pharmacokinetic model selection assisted by machine learning

A fit-for-purpose structural and statistical model is the first major requirement in population pharmacometric model development. In this manuscript we discuss how this complex and computationally intensive task could benefit from supervised machine learning algorithms. We compared the classical pharmacometric approach with two machine learning methods, genetic algorithm and neural networks, in different scenarios based on simulated pharmacokinetic data. Genetic algorithm performance was assessed using a fitness function based on log-likelihood, whilst neural networks were trained using mean square error or binary cross-entropy loss. Machine learning provided a selection based only on statistical rules and achieved accurate selection. The minimization process of genetic algorithm was successful at allowing the algorithm to select plausible models. Neural network classification tasks achieved the most accurate results. Neural network regression tasks were less precise than neural network classification and genetic algorithm methods. The computational gain obtained by using machine learning was substantial, especially in the case of neural networks. We demonstrated that machine learning methods can greatly increase the efficiency of pharmacokinetic population model selection in case of large datasets or complex models requiring long run-times. Our results suggest that machine learning approaches can achieve a first fast selection of models which can be followed by more conventional pharmacometric approaches.

Spoken and Inner Speech-related EEG Connectivity in Different Spatial Direction

Although a significant number of studies have been devoted to the investigation of the electrographic correlates and neurophysiological mechanisms of spoken and inner (imagined) speech, there is a question on which EEG characteristics reflect its content. Considering that speech is a complex cognitive process which requires coordinated activity of a number of cortical structures of the large hemispheres, the EEG coherence values were studied. The values were recorded from 14 channels of 10 young men in the process of real verbalization (spoken speech) and during pronunciation of imagined words designating directions in space (up, down, right, left, forward, backward). It was shown that the level of EEG coherence is generally higher for real verbalization, most significantly at gamma-2-rhythm frequencies (55–70 Hz). Spatial coherence patterns specific to a number of words are formed in the left cerebral hemisphere during imagined utterance of words at gamma-2 frequencies. The application of machine learning and neural network classification has demonstrated a significant similarity of the generated spatial coherent patterns of spoken and inner (imagined) speech. The Multi-layer Perceptron (MLP) neural network classification method has shown the accuracy of word detection in the imagined speech according to brain activity patterns up to 49–61% for 3 classes and 33–40% for 7 classes, with a random recognition rate of 33,3 and 14,2% respectively. The latter indicates a promising application of coherence values and imagined speech denoting directions in space for the development of Brain-computer interfaces (BCIs).

Traditional versus Neural Network Classification Methods for Facial Emotion Recognition

Abstract Facial emotion recognition (FER) is a topic that has gained interest over the years for its role in bridging the gap between Human and Machine interactions. This study explores the potential of real time FER modelling, to be integrated in a closed loop system, to help in treatment of children suffering from Autism Spectrum Disorder (ASD). The aim of this study is to show the differences between implementing Traditional machine learning and Deep learning approaches for FER modelling. Two classification approaches were taken, the first approach was based on classic machine learning techniques using Histogram of Oriented Gradients (HOG) for feature extraction, with a k-Nearest Neighbor and a Support Vector Machine model as classifiers. The second approach uses Transfer Learning based on the popular “Alex Net” Neural Network architecture. The performance of the approaches was based on the accuracy of randomly selected validation sets after training on random training sets of the Oulu-CASIA database. The data analyzed shows that traditional machine learning methods are as effective as deep neural net models and are a good compromise between accuracy, extracted features, computational speed and costs.

Spatiotemporal Information Fusion Method of User and Social Media Activity

Social media check-in data contains a lot of user activity information. Understanding the types of activities and behavior of social media users has important research significance for exploring human mobility and behavior patterns. This paper studies the user activity classification method for Sina Weibo (a very popular Chinese social network service, referred to as “Weibo”), which combines image expression and spatiotemporal data classification technology to realize the identification of the activity behavior represented by the microblog check-in data. Firstly, the user activities represented by the Sina Weibo check-in data are divided into six categories according to POI attribute information: “catering”, “life services”, “campus”, “outdoors”, “entertainment” and “travel”; Then, through the Convolutional Neural Network (CNN) and K-Nearest Neighbor (KNN) classification methods, the image scene information and spatiotemporal information in the check-in data are fused to classify the activity behavior of microblog users. The experimental results show that the proposed method can significantly improve the accuracy of microblog user activity type recognition and provide more effective data support for accurately exploring human behavior activities.

An arrhythmia classification method based on deep learning parallel network model

We propose a parallel neural network classification method to improve the performance of classification of 4 types of arrhythmias: normal beat, supraventricular ectopic beat, ventricular ectopic beat and fused beat. Preprocessing was performed including denoising of ECG signal, segmentation of small-scale heartbeat and large-scale heartbeat and data enhancement. Based on deep learning theory, densely connected convolutional network was applied to improve the limitation of waveform feature extraction, and bidirectional long short-term memory network and efficient channel attention network were combined to enhance the function of time series features and important features of the waveform. The parallel network structure was adopted, and the waveform features of small- scale heartbeat and large-scale heartbeat were input to improve the accuracy of arrhythmia classification at the same time. Softmax was used to carry out the 4 classification tasks of arrhythmia by the parallel network model. The proposed method was verified using MIT-BIH Arrhythmia Database and 3 groups of experiments. The experiments for comparing the classification performance of multiple parallel network models and that of each classification model under different heartbeat input methods showed that the proposed classification model had an overall accuracy, average sensitivity and average specificity of 99.36%, 96.08% and 99.41%, respectively. Convergence performance analysis of the parallel network classification model showed that the training time of the classification model was 41 s. The parallel multi-network classification method can improve the average sensitivity, specificity and training time while maintaining a high overall accuracy, and may thus provide a new technical solution for clinical diagnosis of arrhythmia.

COMPARISON OF METHODS FOR IDENTIFYING USER ROLES IN ONLINE SOCIAL NETWORKS

The development of social media has led to its use as a tool for propaganda and mobilising users to participate in protest movements and political actions aimed at undermining the foundations of society and overthrowing the current government. The impact on social media by the organisers of protest movements has become increasingly targeted and organised. In the context of ensuring public safety and countering destructive influences on social media, it is becoming increasingly important to identify the structure of purposeful impact on social media. Important elements of this structure are the roles played by social network users who participate in the protest movement. The paper presents the data of a survey of the social network VKontakte users in Perm region, who have published protest-related materials during the year 2020. Descriptions of the roles of social network users based on data on their publication activity are presented. Existing methods of identifying the roles of online social network users based on clustering and neural network classification are described. Problems associated with the preparation of datasets for qualitative training of neural networks are indicated. The authors have researched user roles using different clustering methods, and proposed original methods of numerical evaluation of user roles and expert neural network classification of user roles based on artificially synthesized datasets. The results of comparison of different clustering methods, numerical estimation method and expert neural network classification method are presented, their advantages and disadvantages are indicated. High correlation between numerical evaluation method and expert neural network classification method is shown. It is noted that the effectiveness of expert neural network classification of roles of users in social networks is higher than that of various clustering methods. In conclusion, the optimum areas of application of the proposed methods for classifying the roles of social network users are indicated and the directions for further research are outlined.

New neural network classification method for individuals ancestry prediction from SNPs data

Artificial Neural Network (ANN) algorithms have been widely used to analyse genomic data. Single Nucleotide Polymorphisms(SNPs) represent the genetic variations, the most common in the human genome, it has been shown that they are involved in many genetic diseases, and can be used to predict their development. Developing ANN to handle this type of data can be considered as a great success in the medical world. However, the high dimensionality of genomic data and the availability of a limited number of samples can make the learning task very complicated. In this work, we propose a New Neural Network classification method based on input perturbation. The idea is first to use SVD to reduce the dimensionality of the input data and to train a classification network, which prediction errors are then reduced by perturbing the SVD projection matrix. The proposed method has been evaluated on data from individuals with different ancestral origins, the experimental results have shown the effectiveness of the proposed method. Achieving up to 96.23% of classification accuracy, this approach surpasses previous Deep learning approaches evaluated on the same dataset.

Remote sensing-based detection of tea land losses: The case of Lahijan, Iran

Accurate change detection of cropland area and their spatial distributions are important for cropland monitoring, food security, and sustainable development. Tea as a strategic crop in northern Iran, has faced many challenges in the coming decades which has led to a decrease in its area under cultivation. The present study aims to identify changes in tea land area located in central district of Lahijan, Gilan, Iran by using Landsat 5 TM, Landsat 7 ETM+ and Landsat 8 OLI/TIRS images. First, pre-processing stage was performed on satellite images including atmospheric, radiometric and geometric corrections. The supervized neural network classification method was employed, resulting in three classification maps with overall accuracies of 94.82, 95.36 and 97.84% and kappa coefficients of 96.22, 97.91, and 98.42 for 1999, 2011, and 2019, respectively which these were satisfactory. The images were categorized into four different classes, namely tea land, agriculture, forest and urban area. The results of Land Change Modeler showed that during 1999–2019, tea and agricultural lands decreased by 23.46% (2142 ha) and 41.71% (3872 ha). The loss trend of tea area in 1999–2011 has been more remarkable with a decrease of 1943 ha (21.28%). This downward trend continued with a slower slope (2.77% equal to 199 ha). Urbanization with growth rate of 381.71% consumed 2096 ha of tea land. Further, forestland contributed 251 ha to tea land losses. Reversely, the contribution of agricultural land to net change of tea area were positive so that 403 ha in 1999–2011 and 74 ha in 2011–2019 from cropland were converted to tea land. The paper highlights the advantage of digital change detection techniques for policy makers to take appropriate decision to return the situation and to conserve the productive agricultural lands.

Neural network application to petrophysical and lithofacies analysis based on multi-scale data: An integrated study using conventional well log, core and borehole image data

Application of artificial neural network (ANN, e.g., Multi-layer perceptron, MLP) became widespread in the petroleum industry, especially in formation evaluation, reservoir characterization and modeling studies. In this study, the MLP technique is applied to estimate the lithofacies and petrophysical parameters including permeability and rock typing for the upper unit of Abu Roash G reservoir (ARG), Azhar field in Beni Suef basin. Major steps of the workflow started with lithofacies analysis using core and borehole image data to identify five lithofacies as follows; bioturbated mudstone, lenticular-bedded siltstone, massive laminated sandstone, flaser bedded sandstone, and faintly laminated sandstone, which were deposited in tidal conditions. A fully connected neural network was trained and tested on data for lithofacies prediction in wells with no core data and borehole image. Core permeability was estimated by using different variables including effective porosity, bulk density, volume of shale and volume of sand, it was found that neural network prediction for permeability was very satisfactory and can be applied on wider scale in other uncored wells. Rock typing was constructed in cored wells using several methods including porosity-permeability plot integrated with the Winland R35, flow zone indicator-reservoir quality index (FZI-RQI). Neural network classification method was trained and tested on data for rock type estimation in the uncored zones based on calculated petrophysical log data including permeability, flow zone indicator, reservoir quality index and effective porosity. To get more accurate predicted data, the log data was calibrated firstly using porosity and permeability core data.

Atom-substituted tensor dictionary learning enhanced convolutional neural network for hyperspectral image classification

A novel sparse tensor dictionary learning algorithm and a convolutional neural network (CNN) classification method based on this algorithm are proposed for hyperspectral image (HSI) in this paper. The HSI is intuitively represented as a three-dimensional (3-D) cube, and utilizing the joint spatial-spectral information can significantly improve the accuracy of HSI classification. Therefore, the proposed atom-substituted tensor dictionary learning (ASTDL) algorithm utilizes tensor techniques to extract 3-D joint spatial-spectral features from HSI cubes directly. Sparsity constraint is enforced on the coefficient tensors, which obeys the sparsity attribute of HSI. The proposed ASTDL enhanced CNN (ASTDL-CNN) classification method utilizes a two-dimensional (2-D) CNN to extract deep features from the feature tensors obtained by the ASTDL, and to perform the pixel-wise classification. The use of ASTDL, which extracts intrinsic tensor features before CNN, alleviates within-class spatial-spectral variation and reduces the requirement of CNN for the labeled data. Whereas, the CNN works as a 3-D classifier, and provides the final classification results. Besides, we perform the majority vote on the classification map obtained by CNN to refine the classification. The performance of the proposed method is evaluated on three real HSI data sets. The competitive results to the compared state-of-the-art methods demonstrate that the ASTDL-CNN can provide accurate and robust classification results.

A Novel Acoustic Sediment Classification Method Based on the K-Mdoids Algorithm Using Multibeam Echosounder Backscatter Intensity

The modern discrimination of sediment is based on acoustic intensity (backscatter) information from high-resolution multibeam echo-sounder systems (MBES). The backscattering intensity, varying with the angle of incidence, reveals the characteristics of seabed sediment. In this study, we propose a novel unsupervised acoustic sediment classification method based on the K-medoids algorithm using multibeam backscattering intensity data. In this method, we use the Lurton parameters model, which is the relationship between the backscattering intensity and incidence, to obtain the backscattering angle corresponding curve, and we use the genetic algorithm to fit the curve by the least-squares method. After extracting the four relevant parameters of the model when the ideal fitting effect was achieved, we input the characteristic parameters obtained from the fitting to the K-medoids clustering model. To validate the proposed classification method, we compare it with the self-organizing map (SOM) neural network classification method under the same parameter settings. The results of the experiment show that when the seabed sediment category is less than or equal to 3, the results of the K-medoids algorithm and the SOM neural network are approximately identical. As the sediment category increases, the SOM neural network shows instability, and it is impossible to see the clear boundaries of the seabed sediment, while the K-medoids category is 5 and the seabed sediment classification is correct. After comparing with field in situ seabed sediment sampling along the MBES survey line, the sediment classification method based on K-medoids is consistent with the distribution of the field sediment sampling. The classification accuracies for bedrock, sandy clay, and silty sand are all above 90%; those for gravel and clay are nearly 80%, and the overall accuracy reaches 89.7%.

Automatic detection of procedural knowledge in robotic-assisted surgical texts

PurposeThe automatic extraction of knowledge about intervention execution from surgical manuals would be of the utmost importance to develop expert surgical systems and assistants. In this work we assess the feasibility of automatically identifying the sentences of a surgical intervention text containing procedural information, a subtask of the broader goal of extracting intervention workflows from surgical manuals.MethodsWe frame the problem as a binary classification task. We first introduce a new public dataset of 1958 sentences from robotic surgery texts, manually annotated as procedural or non-procedural. We then apply different classification methods, from classical machine learning algorithms, to more recent neural-network approaches and classification methods exploiting transformers (e.g., BERT, ClinicalBERT). We also analyze the benefits of applying balancing techniques to the dataset.ResultsThe architectures based on neural-networks fed with FastText’s embeddings and the one based on ClinicalBERT outperform all the tested methods, empirically confirming the feasibility of the task. Adopting balancing techniques does not lead to substantial improvements in classification.ConclusionThis is the first work experimenting with machine / deep learning algorithms for automatically identifying procedural sentences in surgical texts. It also introduces the first public dataset that can be used for benchmarking different classification methods for the task.

Performance Analysis of Glioma Brain Tumor Segmentation Using CNN Deep Learning Approach

The segmentation of Glioma tumor regions in brain Magnetic Resonance Imaging (MRI) image, using Convolutional Neural Networks (CNN) classification method, is proposed in this paper. The adaptive histogram equalization method is applied on the brain MRI image for enhancing the abnormal pixels with respect to surrounding pixels. This enhanced brain image is transformed into multidirectional scaling image using Gabor transform. Then, features are extracted from this multidirectional scaling image and then these features are trained and classified using CNN deep learning algorithm in order to differentiate the Glioma from normal brain MRI image. Finally, the tumor regions are segmented using morphological operations. The proposed Glioma brain tumor segmentation method using CNN classification approach obtains 96.9% of sensitivity, 99.3% of specificity and 99.2% of accuracy.

An Optimized Neural Network Classification Method Based on Kernel Holistic Learning and Division

An optimized neural network classification method based on kernel holistic learning and division (KHLD) is presented. The proposed method is based on the learned radial basis function (RBF) kernel as the research object. The kernel proposed here can be considered a subspace region consisting of the same pattern category in the training sample space. By extending the region of the sample space of the original instances, relevant information between instances can be obtained from the subspace, and the classifier’s boundary can be far from the original instances; thus, the robustness and generalization performance of the classifier are enhanced. In concrete implementation, a new pattern vector is generated within each RBF kernel according to the instance optimization and screening method to characterize KHLD. Experiments on artificial datasets and several UCI benchmark datasets show the effectiveness of our method.

Recognition of Abnormal Chest Compression Depth Using One-Dimensional Convolutional Neural Networks.

When the displacement of an object is evaluated using sensor data, its movement back to the starting point can be used to correct the measurement error of the sensor. In medicine, the movements of chest compressions also involve a reciprocating movement back to the starting point. The traditional method of evaluating the effects of chest compression depth (CCD) is to use an acceleration sensor or gyroscope to obtain chest compression movement data; from these data, the displacement value can be calculated and the CCD effect evaluated. However, this evaluation procedure suffers from sensor errors and environmental interference, limiting its applicability. Our objective is to reduce the auxiliary computing devices employed for CCD effectiveness evaluation and improve the accuracy of the evaluation results. To this end, we propose a one-dimensional convolutional neural network (1D-CNN) classification method. First, we use the chest compression evaluation criterion to classify the pre-collected sensor signal data, from which the proposed 1D-CNN model learns classification features. After training, the model is used to classify and evaluate sensor signal data instead of distance measurements; this effectively avoids the influence of pressure occlusion and electromagnetic waves. We collect and label 937 valid CCD results from an emergency care simulator. In addition, the proposed 1D-CNN structure is experimentally evaluated and compared against other CNN models and support vector machines. The results show that after sufficient training, the proposed 1D-CNN model can recognize the CCD results with an accuracy rate of more than 95%. The execution time suggests that the model balances accuracy and hardware requirements and can be embedded in portable devices.

Influencing Factors of e‐Commerce Enterprise Development Based on Mobile Computing Big Data Analysis

With the development of increasingly advanced information technology and electronic technology, especially with regard to physical information systems, cloud computing systems, and social services, big data will be widely visible, creating benefits for people and at the same time facing huge challenges. In addition, with the advent of the era of big data, the scale of data sets is getting larger and larger. Traditional data analysis methods can no longer solve the problem of large‐scale data sets, and the hidden information behind big data is digging out, especially in the field of e‐commerce. We have become a key factor in competition among enterprises. We use a support vector machine method based on parallel computing to analyze the data. First, the training samples are divided into several working subsets through the SOM self‐organizing neural network classification method. Compared with the ever‐increasing progress of information technology and electronic equipment, especially the related physical information system finally merges the training results of each working set, so as to quickly deal with the problem of massive data prediction and analysis. This paper proposes that big data has the flexibility of expansion and quality assessment system, so it is meaningful to replace the double‐sidedness of quality assessment with big data. Finally, considering the excellent performance of parallel support vector machines in data mining and analysis, we apply this method to the big data analysis of e‐commerce. The research results show that parallel support vector machines can solve the problem of processing large‐scale data sets. The emergence of data dirty problems has increased the effective rate by at least 70%.

Lung Cancer Diagnosis Based on Chan-Vese Active Contour and Polynomial Neural Network

Lung cancer is of the most serious and common type of cancers. It is usually diagnosed in the final stages which make it hard to treat. Now days, many techniques are to help in the detection of lung cancer, but still there is further need to develop more systems with more improvement and higher accuracy for better detection. In this research work, we introduce a computer-aided detection (CAD) system using computed tomography (CT) scans for nodule classification. The proposed system is divided into four stages that involve image pre-processing by using Gabor filter and Kuwahara filter, image segmentation by applying Chan- Vese active contouring. Feature extraction where features are computed using Discrete Wavelet Transform (DWT) at one, two and three levels of decomposition. After that, 13 features are computed from each wavelet sub-band. As a result the output features are compared and the best output is used to train Polynomial Neural Network (PNN) classification method to classify benign and malignant nodules. The result of the proposed system shows high performance in both the segmentation and the classification with accuracy of 96.66% for the classifying method.