Hierarchical Extreme Learning Machine Research Articles

A Novel Feature Fusion Approach for Classification of Motor Imagery EEG Based on Hierarchical Extreme Learning Machine

A Novel Feature Fusion Approach for Classification of Motor Imagery EEG Based on Hierarchical Extreme Learning Machine

Research on multi-granularity imbalanced knowledge condition monitoring for mechanical equipment based on hierarchical ELM in multi-entropy space

Condition monitoring is essentially a complex multi-classification task that requires struggling with multiple conditions and imbalanced data. To address this issue, this study proposes a hierarchical cognition method that utilizes multi-entropy feature representation to capture the state of equipment under complex working conditions. Firstly, this study proposes a feature modeling method for mechanical equipment based on multi-entropy space, utilizing the cloud model as a space component. Secondly, the study uses multi-entropy feature data to develop a hierarchical Extreme Learning Machine (HELM) fault identification model for complex working conditions and data imbalance inspired by human hierarchical cognitive behavior. The feature data are decomposed into four levels, namely speed, load, abnormality, and fault, respectively. Additionally, a hierarchical feature selection mechanism is designed to improve the efficiency of HELM. Finally, comparative experiments are carried out on two platforms. Experiments show that the If factors of the two datasets in the multi-entropy space reach 0.6888 and 10.4881, respectively. At the same time, in the classification experiment for the imbalanced data, the classification accuracy reaches 90.68 when the fault samples are 10. The results show that the multi-entropy feature created in this study has a specific representation ability for mechanical states. Furthermore, the proposed hierarchical extreme learning method can effectively identify faults under multiple conditions and data imbalance backgrounds.

Efficient and Automatic Breast Cancer Early Diagnosis System Based on the Hierarchical Extreme Learning Machine.

Breast cancer is the leading type of cancer in women, causing nearly 600,000 deaths every year, globally. Although the tumors can be localized within the breast, they can spread to other body parts, causing more harm. Therefore, early diagnosis can help reduce the risks of this cancer. However, a breast cancer diagnosis is complicated, requiring biopsy by various methods, such as MRI, ultrasound, BI-RADS, or even needle aspiration and cytology with the suggestions of specialists. On certain occasions, such as body examinations of a large number of people, it is also a large workload to check the images. Therefore, in this work, we present an efficient and automatic diagnosis system based on the hierarchical extreme learning machine (H-ELM) for breast cancer ultrasound results with high efficiency and make a primary diagnosis of the images. To make it compatible to use, this system consists of PNG images and general medical software within the H-ELM framework, which is easily trained and applied. Furthermore, this system only requires ultrasound images on a small scale, of 28×28 pixels, reducing the resources and fulfilling the application with low-resolution images. The experimental results show that the system can achieve 86.13% in the classification of breast cancer based on ultrasound images from the public breast ultrasound images (BUSI) dataset, without other relative information and supervision, which is higher than the conventional deep learning methods on the same dataset. Moreover, the training time is highly reduced, to only 5.31 s, and consumes few resources. The experimental results indicate that this system could be helpful for precise and efficient early diagnosis of breast cancers with primary examination results.

Improving fault tolerance in diagnosing power system failures with optimal hierarchical extreme learning machine

Improving fault tolerance in diagnosing power system failures with optimal hierarchical extreme learning machine

Differential evolution-based optimized hierarchical extreme learning machines for fault section diagnosis of large-scale power systems

Fault section diagnosis (FSD) is extremely critical to ensure the safe and stable operation of power systems. Neural networks have become popular increasingly in solving this problem and the existing applied ones rely primarily on the traditional BP and RBF neural networks. However, their generalization is weak when diagnosing large-scale power systems. To address this issue, this study suggests a divisional FSD method based on structure adaptive hierarchical extreme learning machines (HELM). To construct an optimal HELM-based sub-module diagnosis model for each subsystem of a large-scale power system, differential evolution is applied to optimize both the number of hidden layer neurons and the regularization factor of HELM simultaneously. The optimal HELM is responsible for completing both the diagnosis of internal power sections of the subsystem and the interim diagnosis of tie lines of adjacent subsystems. Based on the interim diagnosis results, Choquet fuzzy integral is used to fuse them to obtain the final results for the tie lines. Simulation results indicate that the differential evolution algorithm is effective in finding more accurate and compact HELMs with better approximation ability and generalization performance. The proposed FSD method has higher accuracy and fault tolerance to identify various complex fault cases.

Characterization of ash content in wheat flour using data fusion

Ash detection played an important role in wheat flour quality testing. In this work, we proposed a method for the ash content detection in wheat flour, which based on hyperspectral (HSI) method fused with terahertz (THz) technology. Standard Normal Variable Correction (SNV) and Multiple Scattering Correction (MSC) were used for the pretreatment of HSI and THz data. Moreover, feature wavelengths were extracted by the Successive Projections Algorithm (SPA) and Feature weighting algorithms (Relieff), respectively. Hierarchical Extreme Learning Machine (H-ELM) model were used for the analysis of two spectral data, which combined with data fusion strategy. The prediction coefficient of determination (r2) and the root mean square error of prediction (RMSEP) were employed to evaluate the performance of the model effectively. Results suggested that the model accuracy of data fusion was better than that of single spectrum model. Compared to other fused method, this model obtained better results at r2 = 0.989 and RMSEP = 0.015, which based on the SPA-selected HSI feature data fused with the Relieff-selected THz feature data. This study demonstrates that data fusion analysis offers a way to detect ash content in wheat flour, which based on HSI method and THz technology.

Hierarchical extreme puzzle learning machine-based emotion recognition using multimodal physiological signals

Detection of exact emotions through multi-modal physiological signals provides relevant information for different processes. Numerous computational approaches have been presented for the precise analysis of emotion types. But due to some problems like ruined signal quality, increased time consumption, and the necessity of high storage space, classification accuracy’s efficiency worsens. Hence, this research classified multi-modal physiological signals based on machine and deep learning (DL) models. The proposed work implements the Hierarchical Extreme Puzzle Learning Machine (HEPLM) approach to classify the actual output of embedded emotions. The proposed work comprises four steps: pre-processing, signal-to-image conversion, feature extraction, and classification. Pre-processing is carried out using Savitzky-Golay smoothing filtering (SGF) for the removal of noise and to increase signal quality. Hybrid wavelet scattering with Synchro squeezing Wavelet Transform approach converts the signal into an image. In feature extraction process, the valuable features are extracted using ResNet-152 and the Inception v3 model, whereas the features are combined through an ensemble approach. HEPLM is used in the final classification process, combining Puzzle Optimization Algorithm (POA) and Hierarchical Extreme Learning Machine (HELM) to reduce feature dimensionality and improve classification accuracy. The dataset adopted in the proposed work is Wearable Stress and Affect Detection (WESAD) to collect multi-modal physiological signals. The presentation of the projected work is assessed with metrics like accuracy, recall, precision, F1 score, kappa, and so on. The proposed effort demonstrates better results of emotion classification when compared to the existing methods by holding 96.29% of accuracy.

Pathological voice classification based on multi-domain features and deep hierarchical extreme learning machine

The intelligent data-driven screening of pathological voice signals is a non-invasive and real-time tool for computer-aided diagnosis that has attracted increasing attention from researchers and clinicians. In this paper, the authors propose multi-domain features and the hierarchical extreme learning machine (H-ELM) for the automatic identification of voice disorders. A sufficient number of sensitive features are first extracted from the original voice signal through multi-domain feature extraction (i.e., features of the time domain and the sample entropy based on ensemble empirical mode decomposition and gammatone frequency cepstral coefficients). To eliminate redundancy in high-dimensional features, neighborhood component analysis is then applied to filter out sensitive features from the high-dimensional feature vectors to improve the efficiency of network training and reduce overfitting. The sensitive features thus obtained are then used to train the H-ELM for pathological voice classification. The results of the experiments showed that the sensitivity, specificity, F1 score, and accuracy of the H-ELM were 99.37%, 98.61%, 99.37%, and 98.99%, respectively. Therefore, the proposed method is feasible for the initial classification of pathological voice signals.

Multilayer Fisher extreme learning machine for classification

As a special deep learning algorithm, the multilayer extreme learning machine (ML-ELM) has been extensively studied to solve practical problems in recent years. The ML-ELM is constructed from the extreme learning machine autoencoder (ELM-AE), and its generalization performance is affected by the representation learning of the ELM-AE. However, given label information, the unsupervised learning of the ELM-AE is difficult to build the discriminative feature space for classification tasks. To address this problem, a novel Fisher extreme learning machine autoencoder (FELM-AE) is proposed and is used as the component for the multilayer Fisher extreme leaning machine (ML-FELM). The FELM-AE introduces the Fisher criterion into the ELM-AE by adding the Fisher regularization term to the objective function, aiming to maximize the between-class distance and minimize the within-class distance of abstract feature. Different from the ELM-AE, the FELM-AE requires class labels to calculate the Fisher regularization loss, so that the learned abstract feature contains sufficient category information to complete classification tasks. The ML-FELM stacks the FELM-AE to extract feature and adopts the extreme leaning machine (ELM) to classify samples. Experiments on benchmark datasets show that the abstract feature extracted by the FELM-AE is more discriminative than the ELM-AE, and the classification results of the ML-FELM are more competitive and robust in comparison with the ELM, one-dimensional convolutional neural network (1D-CNN), ML-ELM, denoising multilayer extreme learning machine (D-ML-ELM), multilayer generalized extreme learning machine (ML-GELM), and hierarchical extreme learning machine with L21‑norm loss and regularization (H-LR21-ELM).

Non-invasive diagnosis of fetal arrhythmia based on multi-domain feature and hierarchical extreme learning machine

Heart disease is one of the major causes of affecting the health of newborns. Detecting the presence or potential heart disease of the fetus as soon as possible, and adopting relevant treatment plans in a timely manner, which has a profound impact on doctors and patients. This study aims to develop an accurate screening method with arrhythmia (ARR) to assist physicians to further diagnose whether them have heart disease. Therefore, this paper proposes a multi-domain feature extraction technique and a hierarchical extreme learning machine (H-ELM) network for the prediction of fetal ARR. Firstly, the multi-domain feature extraction technology is used to extract abundant high-dimensional feature for representing the original signal. Secondly, neighborhood component analysis (NCA) is used to screen the sensitive features from the high dimensional feature vectors. Then, the obtained sensitive features are input into stacked extreme learning machine sparse autoencoder (ELM-SAE), which extract high-level fusion features by layer-by-layer unsupervised learning manner. Finally, an original ELM was connected on the end of the ELM-SAE network for the prediction of fetal ARR. The experimental results illustrate that the proposed method can achieve sensitivity of 99.11%, specificity of 93.91%, precision of 93.52%, and accuracy of 96.33%. Furthermore, the proposed method comprehensive performance outperforms the compared models. Therefore, the proposed method can be effectively used for the prediction of fetal ARR, and with the continuous improved the research, it is expected to be considered as an auxiliary diagnostic tool for physicians in the future.

Construction of Data Driven Decomposition Based Soft Sensors with Auto Encoder Deep Neural Network for IoT Healthcare Applications

The architecture of IoT healthcare is motivated towards the data-driven realization and patient-centric health models, whereas the personalized assistance is provided by deploying the advanced sensors. According to the procedures in surgery, in the emergency unit, the patients are monitored till they are stable physically and then shifted to ward for further recovery and evaluation. Normally evaluation done in ward doesn’t suggest continuous parameters monitoring for physiological condition and thus relapse of patients are common. In real-time healthcare applications, the vital parameters will be estimated through dedicated sensors, that are still luxurious at the present situation and highly sensitive to harsh conditions of environment. Furthermore, for real-time monitoring, delay is usually present in the sensors. Because of these issues, data-driven soft sensors are highly attractive alternatives. This research is motivated towards this fact and Auto Encoder Deep Neural Network (AutoEncDeepNN) is proposed depending on Health Framework in the internet assisting the patients with trigger-based sensor activation model to manage master and slave sensors. The advantage of the proposed method is that the hidden information are mined automatically from the sensors and high representative features are generated by multiple layer’s iteration. This goal is consistently achieved and thus the proposed model outperforms few standard approaches which are considered like Hierarchical Extreme Learning Machine (HELM), Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM). It is found that the proposed AutoEncDeepNN method achieves 94.72% of accuracy, 41.96% of RMSE, 34.16% of RAE and 48.68% of MAE in 74.64 ms.

Hierarchical Extreme Learning Machine Enabled Dynamic Line Rating Forecasting

Dynamic line rating (DLR) is a technology introduced to rectify an overhead line's current carrying capacity based on climatic conditions. Reducing congestion costs, increasing the penetration of renewable energy, and network stability are the most important benefits of this modification. The DLR forecasting is a significant issue and can be very useful in network management decisions for real-time and day-ahead. So far, however, very few studies have talked about the literature on day-ahead DLR forecasting. This article develops various DLR forecasting models based on historical meteorological data and machine learning techniques. Accordingly, multilayer perceptron, group method data handling, support vector regression, back-propagation neural network, extreme learning machine (ELM), and hierarchical ELM (H-ELM) techniques compete for short-term DLR forecasting of two 400 kV overhead transmission lines, installed in the Khaf and Ghadamgah regions in Iran. The meteorological parameters such as air temperature, wind speed, wind direction, and solar radiation in the studied areas are utilized as input variables for selected methods. Different evaluation indicators assessments the forecasting results of each technique. Finally, evaluations emphasize the effectiveness and capability of the H-ELM method with the lowest forecasting error values and the highest correlation coefficient value compared to other methods. Experiments exposed the DLR forecasting models' generalizability at various points of the line without the extension of monitoring devices and communication infrastructures.

Tidal current prediction based on a hybrid machine learning method

Accurate and reliable tidal current forecasting plays a key role in application of grid-connected tidal current power generation systems. Hence, a novel tidal current prediction method based on a hybrid machine learning method is proposed in this paper. Compared with previous tidal current prediction method, the proposed tidal current prediction method considers the influence of turbulence on multi-layer tidal velocity difference. A hybrid machine learning method based on hierarchical extreme learning machine(H-ELM) and long short-term memory (LSTM) is adopted to reduce effect of the turbulent flow, which enables to increase the accuracy of the prediction result. The tidal data that collected by using acoustic doppler current profiler (ADCP) in Zhejiang province China is used to validate the effectiveness of the proposed method. The results of the case study show that the proposed method has higher accuracy than results from standard harmonic analysis using UTide (a freely available academic research product).

A new feature selection approach for driving fatigue EEG detection with a modified machine learning algorithm

This study aims to identify new electroencephalography (EEG) features for the detection of driving fatigue. The most common EEG feature in driving fatigue detection is the power spectral density (PSD) of five frequency bands, i.e., alpha, beta, gamma, delta, and theta bands. PSD has proved to be useful, however its flaw is that it covers much implicit information of the time domain. In this study we propose a new approach, which combines ensemble empirical mode decomposition (EEMD) and PSD, to explore new EEG features for driving fatigue detection. Through EEMD we get a series of intrinsic mode function (IMF) components, from which we can extract PSD features. We used six features to compare with the proposed features, including the PSD of five frequency bands, PSD of empirical mode decomposition (EMD)-IMF components, PSD, permutation entropy (PE), sample entropy (SE), and fuzzy entropy (FE) of EEMD-IMF components, and common spatial pattern. Feature overlap ratio and multiple machine learning methods were applied to evaluate these feature extraction approaches. The results show that the classification accuracy and overlap ratio of experiments based on IMF's energy spectrum is far superior to other features. Through channel optimization and a comparison of accuracy, we conclude that our new feature selection approach has a better performance based on the modified hierarchical extreme learning machine algorithm with Particle Swarm Optimization (PSO–H-ELM) classifier, which has the highest average accuracy of 97.53%.

Estimation of rip density on intermediate beaches using an extreme learning machine model

Rip currents are one of the most well-known and risky flows in coastal areas, and given the limitations and difficulties in field data collection, predicting the varied behavior of these flows is of particular importance. In this study, based on the design of the intermediate beaches, the wave and bed parameters in the coastal zone were extracted as dimensionless ratios of Froude number, wave height, surf zone width and rip channel width by Mike 21/3 numerical model, and were used as input for different approaches of extreme learning machine (ELM) and artificial neural network (ANN). Then, while examining the different beach states over time, the optimum values for statistical criteria in different models were investigated and compared. Next, the effect of each of the available parameters on the rip density was considered, and the results obtained by different models in different climates of the waves were compared with the field results, which showed high level of agreement. The main results of the study revealed that the speed of implementation of different test, training and validation phases in ELM model was 1.5 times faster than ANN model on average. Furthermore, it was shown that the different percentages of the rip density values observed by the numerical model and predicted by different methods of the ELM model were always lower than those of the ANN model. Also, the dispersion of the predicted data by different models of the extreme learning machine, especially hierarchical extreme learning machine (HELM), was less than the dispersion of the data obtained by the other models, which makes the HELM model superior to other soft computing methods in predicting the rip density on intermediate beaches.

Manifold semi-supervised learning for aluminum electrolysis temperature identification based on regularized hierarchical extreme learning machine

The aim of this article is to develop a soft approach for a real-time cell temperature prediction in the aluminum electrolysis reduction. Under the limited labeled data constraint, Laplacian semi-supervised learning methods, which can fully utilize the underlying structure of the data distribution and further extract information contained in all available data, has recently received extensive attention in the field of soft sensor modeling. Since the Laplacian underlying manifold is a constant, it remains a challenging task to improve the extrapolating ability for the case that only a few labeled samples are available. This study presents a soft modeling method based on a semi-supervised deep learning structure, which was developed from the hierarchical autoencoders with extreme learning machine. Furthermore, a Laplacian–Hessian semi-supervised extreme learning machine is built to learn all the geometric distribution information. The Laplacian–Hessian semi-supervised extreme learning machine method is applied to estimate the cell temperature in an aluminum reduction process. The experimental results demonstrate the performance and robustness of the proposed algorithm are superior to those of the existing state-of-the-art methods.

Affinity propagation clustering-aided two-label hierarchical extreme learning machine for Wi-Fi fingerprinting-based indoor positioning

Affinity propagation clustering-aided two-label hierarchical extreme learning machine for Wi-Fi fingerprinting-based indoor positioning

Online Estimating State of Health of Lithium-Ion Batteries Using Hierarchical Extreme Learning Machine

Battery state-of-health (SoH) monitoring is of great importance to ensure the safety and reliability of battery systems. This study proposed an innovative SoH estimation method using hierarchical extreme learning machine (HELM) to improve the estimation robustness and accuracy without the complex parameter model was directly applied to establish the HELM-oriented online SoH estimation framework. First, the increase in mean ohmic resistance was constructed as a novel health indicator (HI) to characterize battery aging. Then, the HI was adopted for offline training to build an HELM model, which captures the underlying correlation between the extracted HI and capacity degradation. Finally, the datasets of four batteries at three different temperatures with dynamic loading profiles were used for validation. The results show that the SoH estimation errors are no more than 1.5%, while the training and estimation datasets are from the same temperature; when the SoH estimation is conducted at different temperatures, the maximum error is only 3.36%. The results indicate that the proposed method had good generalization and reliability for SoH estimation, which is applicable for dynamic scenarios with different temperatures.

Classification of Parkinson’s disease patients based on spectrogram using local binary pattern descriptors

Extreme learning machine is an algorithm that has shown a good performance facing classification and regression problems. It has gained great acceptance by the scientific community due to the simplicity of the model and its sola great generalization capacity. This work proposes the use of extreme learning machine neural networks to carry out the classification between Parkinson’s disease patients and healthy individuals. The descriptor used corresponds to the feature vector generated applying the local binary Pattern algorithm to the grayscale spectrograms. The spectrograms are obtained from the audio signal samples from the considered repository. Experiments are conducted with single hidden layer and multilayer extreme learning machine networks comparing the results of each structure. Results show that hierarchical extreme learning machine with three hidden layers has a better general performance over multilayer extreme learning machine networks and a single hidden layer extreme learning machine. The rate of success obtained is within the ranges presented in the literature. However, the hierarchical network training time is considerably faster compared to multilayer networks of three or two hidden layers.

Change Detection for Heterogeneous Remote Sensing Images with Improved Training of Hierarchical Extreme Learning Machine (HELM)

To solve the problems of susceptibility to image noise, subjectivity of training sample selection, and inefficiency of state-of-the-art change detection methods with heterogeneous images, this study proposes a post-classification change detection method for heterogeneous images with improved training of hierarchical extreme learning machine (HELM). After smoothing the images to suppress noise, a sample selection method is defined to train the HELM for each image, in which the feature extraction is respectively implemented for heterogeneous images and the parameters need not be fine-tuned. Then, the multi-temporal feature maps extracted from the trained HELM are segmented to obtain classification maps and then compared to generate a change map with changed types. The proposed method is validated experimentally by using one set of synthetic aperture radar (SAR) images obtained from Sentinel-1, one set of optical images acquired from Google Earth, and two sets of heterogeneous SAR and optical images. The results show that compared to state-of-the-art change detection methods, the proposed method can improve the accuracy of change detection by more than 8% in terms of the kappa coefficient and greatly reduce run time regardless of the type of images used. Such enhancement reflects the robustness and superiority of the proposed method.