Extreme Learning Machine Classifier Research Articles

Stacked autoencoder based deep random vector functional link neural network for classification

Extreme learning machine (ELM), which can be viewed as a variant of Random Vector Functional Link (RVFL) network without the input–output direct connections, has been extensively used to create multi-layer (deep) neural networks. Such networks employ randomization based autoencoders (AE) for unsupervised feature extraction followed by an ELM classifier for final decision making. Each randomization based AE acts as an independent feature extractor and a deep network is obtained by stacking several such AEs. Inspired by the better performance of RVFL over ELM, in this paper, we propose several deep RVFL variants by utilizing the framework of stacked autoencoders. Specifically, we introduce direct connections (feature reuse) from preceding layers to the fore layers of the network as in the original RVFL network. Such connections help to regularize the randomization and also reduce the model complexity. Furthermore, we also introduce denoising criterion, recovering clean inputs from their corrupted versions, in the autoencoders to achieve better higher level representations than the ordinary autoencoders. Extensive experiments on several classification datasets show that our proposed deep networks achieve overall better and faster generalization than the other relevant state-of-the-art deep neural networks.

FPGA Implementation of Weighted Online Sequential Extreme Learning Machine for Data Classification

To design an efficient embedded module field-programmable gate array (FPGA) plays significant role. FPGA, a high speed reconfigurable hardware platform has been used in various field of research to produce the throughput efficiently. A now-a-days artificial neural network (ANN) is the most prevalent classifier for many analytical applications. In this paper, weighted online sequential extreme learning machine (WOS-ELM) classifier is presented and implemented in hardware environment to classify the different real-world bench-mark datasets. The faster learning speed, remarkable classification accuracy, lesser hardware resources, and short-event detection time, aid the hardware implementation of WOS-ELM classifier to design an embedded module. Finally, the developed hardware architecture of the WOS-ELM classifier is implemented on a high speed reconfigurable Xilinx Virtex (ML506) FPGA board to demonstrate the feasibility, effectiveness, and robustness of WOS-ELM classifier to classify the data in real-time environment.

ELM-Based Large-Scale Genetic Association Study via Statistically Significant Pattern

Genetic association study (GAS) is a promising tool for detecting and analyzing the cause of complex diseases. The extreme learning machine (ELM) has been successfully applied in a variety of research fields. Yet, as a black box method, it could not measure up to the task of GAS by itself, because it cannot tell us what causes the diseases, which is very crucial to the biologists. In this paper, we propose an ELM-based statistically significant pattern classification framework, which combines ELM with feature vector-based methods to solve the GAS problem efficiently and effectively. In particular: 1) a statistically significant pattern considering in terms of both family wise error rate (FWER) and false discovery rate (FDR) is proposed to control false positives in multiple hypothesis tests, which is necessary in GAS, but ignored by most of the existing methods; 2) an upper bound of the significance of a pattern is deduced to speed up FWER-constrained statistically significant pattern mining in a row enumeration way. Further, a space-effective grid index is devised to dramatically improves the efficiency of FDR-constrained pattern discovery; and 3) an ELM classifier is constructed based on the significant patterns. Comprehensive empirical studies on four real genotype datasets demonstrate much higher efficiency and effectiveness of our proposed framework with respect to the compared methods.

A fused CNN model for WBC detection with MRMR feature selection and extreme learning machine

White blood cell (WBC) test is used to diagnose many diseases, particularly infections, ranging from allergies to leukemia. A physician needs clinical experience to detect and classify the amount of WBCs in human blood. WBCs are divided into four subclasses: eosinophils, lymphocytes, monocytes, and neutrophils. In the present study, pre-trained architectures, namely AlexNet, VGG-16, GoogleNet, and ResNet, were used as feature extractors. The features obtained from the last fully connected layers of these architectures were combined. Efficient features were selected using the minimum redundancy maximum relevance method. Finally, unlike classical convolutional neural network (CNN) architectures, the extreme learning Machine (ELM) classifier was used in the classification stage thanks to the efficient features obtained from CNN architectures. Experimental results indicated that efficient CNN features yielded satisfactory results in a shorter execution time via ELM classification with an accuracy rate of 96.03%.

Coarse-to-fine SAR image change detection method

ABSTRACTIn this paper, in order to obtain high-quality training samples and improve the accuracy of the synthetic aperture radar (SAR) image change detection, we propose a coarse-to-fine SAR image change detection method. In the coarse change detection stage, we construct the difference image (DI) by the use of the wavelet frequency difference (WFD) method, and obtain the DI saliency map by the use of quaternion Fourier transform (QFT). At the same time, the noise in the non-salient regions is suppressed. Finally, the coarse change detection map obtained by selecting a threshold for the DI saliency map is pre-classified (changed pixels, unchanged pixels, undetermined pixels) by the fuzzy c-means (FCM) clustering algorithm. In the fine change detection phase, the neighbourhood features of the changed pixels and the unchanged pixels in the coarse change map are extracted and used as reliable samples for extreme learning machine (ELM) training. The trained ELM classifier is then used to perform change detection on the coarse change detection map, to obtain the final change detection map. Experiments on two real SAR datasets show that the proposed method can not only obtain reliable training samples, but it can also result in a significant improvement in change detection performance.

An Integrated Methodology for Rule Extraction from ELM-Based Vacuum Tank Degasser Multiclassifier for Decision-Making

The present work proposes an integrated methodology for rule extraction in a vacuum tank degasser (VTD) for decision-making purposes. An extreme learning machine (ELM) algorithm is established for a three-class classification problem according to an end temperature of liquid steel that is higher than its operating restriction, within the operation restriction and lower than the operating restriction. Based on these black-box model results, an integrated three-step approach for rule extraction is constructed to interpret the understandability of the proposed ELM classifier. First, the irrelevant attributes are pruned without decreasing the classification accuracy. Second, fuzzy rules are generated in the form of discrete input attributes and the target classification. Last but not the least, the rules are refined by generating rules with continuous attributes. The novelty of the proposed rule extraction approach lies in the generation of rules using the discrete and continuous attributes at different stages. The proposed method is analyzed and validated on actual production data derived from a No.2 steelmaking workshop in Baosteel. The experimental results revealed that the extracted rules are effective for the VTD system in classifying the end temperature of liquid steel into high, normal, and low ranges. In addition, much fewer input attributes are needed to implement the rules for the manufacturing process of VTD. The extracted rules serve explicit instructions for decision-making for the VTD operators.

Environmental sound classification using optimum allocation sampling based empirical mode decomposition

Automatic environmental sound classification (ESC) is prominent in various fields like robotics, security, and crime investigation. In this paper, optimum allocation sampling (OAS)-based empirical mode method (EMD) is proposed for automatic ESC. The OAS provides the reduced homogeneous length sequence of each long length sound signal, which is further decomposed into band-limited intrinsic mode functions (IMFs) using EMD. The features namely approximate entropy (AE), permutation entropy (PE), log energy entropy (LE), interquartile range (IQR), and zero cross rate (ZCR) are extracted from the IMFs. The OAS-EMD based features used as input to multi-class least squares support vector machine (MC-LS-SVM) and extreme learning machine (ELM) classifiers for evaluation the performance of proposed method. Experimental results show an accuracy of 87.25% and 77.61% with MC-LS-SVM and ELM classifiers, respectively.

Adaptive Particle Swarm Optimization based Credentialed Extreme Learning Machine Classifier (APSO-CELMC) for High Dimensional Datasets

Data mining is a key research field in the computer science research arena. Feature selection is performed once the dataset got cleansed. Optimization algorithms are considered to be helpful for the feature selection task. Also the obtained suitable features will contribute considerably for the classifier. Machine learning classifiers are comparatively performing better than that of traditional data mining classification algorithms. In this part of research work an adaptive particle swarm optimization algorithm is employed in order to perform feature selection task. Extreme learning machine classifier is added with credential weights. Twenty datasets are taken for performance analysis. From the obtained results it is evident that Adaptive Particle Swarm Optimization based Credentialed Extreme Learning Machine Classifier (APSO-CELMC) performs better in terms of predictive accuracy and time taken for classification.

Label Propagation Ensemble for Hyperspectral Image Classification

The imbalance between limited labeled pixels and high dimensionality of hyperspectral data can easily give rise to Hughes phenomenon. Semisupervised learning (SSL) methods provide promising solutions to address the aforementioned issue. Graph-based SSL algorithms, also called label propagation methods, have obtained increasing attention in hyperspectral image (HSI) classification. However, the graphs constructed by utilizing the geometrical structure similarity of samples are unreliable due to the high dimensionality and complexity of the HSIs, especially for the case of very limited labeled pixels. Our motivation is to construct label propagation ensemble (LPE) model, then use the decision fusion of multiple label propagations to obtain pseudolabeled pixels with high classification confidence. In LPE, random subspace method is introduced to partition the feature space into multiple subspaces, then several label propagation models are constructed on corresponding subspaces, finally the results of different label propagation models are fused at decision level, and only the unlabeled pixels whose label propagation results are the same will be assigned with pseudolabels. Meanwhile extreme learning machine classifiers are trained on the labeled and pseudolabeled samples during the iteration. Compared with traditional label propagation methods, our proposed method can deal with the situation of very limited labeled samples by providing pseudolabeled pixels with high classification confidence, consequently, the accurate base classifiers are obtained. To demonstrate the effectiveness of the proposed method, LPE is compared with several state-of-the-art methods on four hyperspectral datasets. In addition, the method that only use label propagation is investigated to show the importance of ensemble technique in LPE. The experimental results demonstrate that the proposed method can provide competitive solution for HSI classification.

Detection and classification of microcalcification from digital mammograms with firefly algorithm, extreme learning machine and non‐linear regression models: A comparison

AbstractIn this study, abnormalities in medical images are analysed using three classifiers, and the results are compared. Breast cancer remains a major public health problem among women worldwide. Recently, many algorithms have evolved for the investigation of breast cancer diagnosis through medical imaging. A computer‐aided microcalcification detection method is proposed to categorise the nature of breast cancer as either benign or malignant from input mammogram images. The standard mammogram image corpus, the Mammogram Image Analysis Society database is utilised, and feature extraction is performed using five different wavelet families at level 4 and level 6 decomposition. The work is accomplished through firefly algorithm (FA), extreme learning machine (ELM) and least‐square‐based non‐linear regression (LSNLR) classifiers. The performance of the classifiers is compared by benchmark metrics, such as total error rate, specificity, sensitivity, area under the receiver operating characteristic curve, precision, F1 score and the Matthews correlation coefficient. As validation of the classifier results, a kappa analysis is included to determine the agreement among classifiers. The LSNLR classifier attains a 3% to 7% improvement in average accuracy compared with the average classification accuracy of the FA (86.75%) and ELM (90.836%) classifiers.

A novel robust feature extraction with GSO-optimized extreme learning for age-invariant face recognition

ABSTRACTThis paper presents a novel age function modelling technique on the basis of the fusion of local features obtained by local texture descriptors. Initially, image normalization is performed and a feature extraction process is carried out. The age estimation performances of new texture descriptors Local Phase Quantization, Weber Local Descriptor and the familiar texture descriptor Local Binary Patterns, which are not examined thoroughly for age estimation modelling, are analysed in this paper. Then the feature fusion process is taken place for investigating the age estimation precisions of various concatenation of the local texture descriptors. By using PCA, dimensionality reduction is implemented for reducing the dimensions of the images. Extreme Learning Machine (ELM) classifier is applied to evaluate the output images for the corresponding input images. Because of the mild optimization restrictions, ELM can be simple for execution and normally attains the finer generalization performance. The outcomes display that, when compared with the earlier techniques, the age function modelling accuracy of the developed system is better.

STREAMFLOW AND SOIL MOISTURE FORECASTING WITH HYBRID DATA INTELLIGENT MACHINE LEARNING APPROACHES: CASE STUDIES IN THE AUSTRALIAN MURRAY–DARLING BASIN

For a drought-prone agricultural nation such as Australia, hydro-meteorological imbalances and increasing demand for water resources are immensely constraining terrestrial water reservoirs and regional-scale agricultural productivity. Two important components of the terrestrial water reservoir i.e., streamflow water level (SWL) and soil moisture (SM), are imperative both for agricultural and hydrological applications. Forecasted SWL and SM can enable prudent and sustainable decisionmaking for agriculture and water resources management. To feasibly emulate SWL and SM, machine learning data-intelligent models are a promising tool in today’s rapidly advancing data science era. Yet, the naturally chaotic characteristics of hydro-meteorological variables that can exhibit non-linearity and non-stationarity behaviors within the model dataset, is a key challenge for non-tuned machine learning models. Another important issue that could confound model accuracy or applicability is the selection of relevant features to emulate SWL and SM since the use of too fewer inputs can lead to insufficient information to construct an accurate model while the use of an excessive number and redundant model inputs could obscure the performance of the simulation algorithm. This research thesis focusses on the development of hybridized dataintelligent models in forecasting SWL and SM in the upper layer (surface to 0.2 m) and the lower layer (0.2–1.5 m depth) within the agricultural region of the Murray-Darling Basin, Australia. The SWL quantifies the availability of surface water resources, while, the upper layer SM (or the surface SM) is important for surface runoff, evaporation, and energy exchange at the Earth-Atmospheric interface. The lower layer (or the root zone) SM is essential for groundwater recharge purposes, plant uptake and transpiration. This research study is constructed upon four primary objectives designed for the forecasting of SWL and SM with subsequent robust evaluations by means of statistical metrics, in tandem with the diagnostic plots of observed and modeled datasets. The first objective establishes the importance of feature selection (or optimization) in the forecasting of monthly SWL at three study sites within the Murray-Darling Basin. Artificial neural network (ANN) model optimized with iterative input selection (IIS) algorithm named IIS-ANN is developed whereby the IIS algorithm achieves feature optimization. The IIS-ANN model outperforms the standalone models and a further hybridization is performed by integrating a nondecimated and advanced maximum overlap discrete wavelet transformation (MODWT) technique. The IIS selected inputs are transformed into wavelet subseries via MODWT to unveil the embedded features leading to IIS-W-ANN model. The IIS-W-ANN outperforms the comparative IIS-W-M5 Model Tree, IIS-based and standalone models. In the second objective, improved self-adaptive multi-resolution analysis (MRA) techniques, ensemble empirical mode decomposition (EEMD) and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) are utilized to address the non-stationarity issues in forecasting monthly upper and lower layer soil moisture at seven sites. The SM time-series are decomposed using EEMD/CEEMDAN into respective intrinsic mode functions (IMFs) and residual components. Then the partial-auto correlation function based significant lags are utilized as inputs to the extreme learning machine (ELM) and random forest (RF) models. The hybrid EEMD-ELM yielded better results in comparison to the CEEMDAN-ELM, EEMD-RF, CEEMDAN-RF and the classical ELM and RF models. Since SM is contingent upon many influential meteorological, hydrological and atmospheric parameters, for the third objective sixty predictor inputs are collated in forecasting upper and lower layer soil moisture at four sites. An ANN-based ensemble committee of models (ANN-CoM) is developed integrating a two-phase feature optimization via Neighborhood Component Analysis based feature selection algorithm for regression (fsrnca) and a basic ELM. The ANN-CoM shows better predictive performance in comparison to the standalone second order Volterra, M5 Model Tree, RF, and ELM models. In the fourth objective, a new multivariate sequential EEMD based modelling is developed. The establishment of multivariate sequential EEMD is an advancement of the classical single input EEMD approach, achieving a further methodological improvement. This multivariate approach is developed to allow for the utilization of multiple inputs in forecasting SM. The multivariate sequential EEMD optimized with cross-correlation function and Boruta feature selection algorithm is integrated with the ELM model in emulating weekly SM at four sites. The resulting hybrid multivariate sequential EEMD-Boruta-ELM attained a better performance in comparison with the multivariate adaptive regression splines (MARS) counterpart (EEMD-Boruta-MARS) and standalone ELM and MARS models. The research study ascertains the applicability of feature selection algorithms integrated with appropriate MRA for improved hydrological forecasting. Forecasting at shorter and near-real-time horizons (i.e., weekly) would help reinforce scientific tenets in designing knowledge-based systems for precision agriculture and climate change adaptation policy formulations.

An image classification framework exploring the capabilities of extreme learning machines and artificial bee colony

A hybridized image classification strategy is proposed based on discrete wavelet transform, artificial bee colony (ABC) and extreme learning machine (ELM). The proposed methodology works in three phases: (a) in preprocessing phase, images are decomposed and features are extracted from images using bi-orthogonal wavelet functions; (b) secondly, modified ABC (MABC) optimization algorithm is proposed to determine the optimal parameters such as hidden layer weights and biases to be used by ELM for classification; (c) the ELM in the third phase has been trained and tested with three brain image datasets for different diseases along with normal brain images. The performance recognition of the proposed MABC-ELM in terms of accuracy, rate of per-image classification and speedup has been made with variants of ELM such as ELM, ABC-ELM and MABC-ELM and also with MLPNN, naive Bayesian, linear regression classifiers. Finally, the percentage of accuracy observed by the proposed MABC-ELM, for acute stroke-speech arrest, glioma and multiple sclerosis datasets, is 90%, 90% and 100% with eight hidden nodes in the ELM architecture, and it can be concluded that MABC-ELM gives better generalization performance, more compact network architecture and the hybridization of ELM with modified ABC is worth investigated.

A mechanical data analysis using kurtogram and extreme learning machine

Today’s industry demands precise functioning and zero failure of rotating machinery (RM) to avoid disastrous accidents as well as financial losses. Rolling element bearings (REBs) are the heart of RM. Therefore, as early as possible to provide the significant time for maintenance planning, an intelligent diagnosis of REB fault is a critical and challenging task. Thus, this paper presents an efficient method for fault diagnosis. The proposed method mainly consists of two consecutive units: (1) generation of kurtogram of raw vibration signal and (2) training of extreme learning machine (ELM) classifier using kurtogram. Kurtogram has a distinct capability to represent the hidden non-stationary components of a raw signal. Therefore, it is considered as a unique feature vector for fault classification. ELM is a well-organized fast learning method proposed by Huang et al. and showed that it is better than traditional learning algorithms. However, one of the open issues of ELM is to design compact-size ELM architecture by preserving the accuracy of the solution. Thus, improved random increment ELM is proposed in this paper. Initially, it randomly adds the nodes to network architecture to rapidly reduce the residual error up to the predefined threshold and then sequentially adds the nodes to the network architecture for further reducing the residual error. Performance of the proposed routine is evaluated by REB vibration data: artificially generated vibration data and Case Western Reserve University bearing data. The experimental study reveals the classification accuracy of the proposed approach with both the datasets for various faults and also compared with existing methods.

Flexible Analytic Wavelet Transform Based Features for Physical Action Identification Using sEMG Signals

ObjectivesElectromyography (EMG) is recording of the electrical activity produced by skeletal muscles. The classification of the EMG signals for different physical actions can be useful in restoring some or all of the lost motor functionalities in these individuals. Accuracy in classifying the EMG signal indicates efficient control of prosthesis. Material and methodsThe flexible analytic wavelet transform (FAWT) is used for classification of surface electromyography (sEMG) signals for identification of physical actions. FAWT is an efficient method for decomposition of sEMG signal into eight sub-bands, features namely neg-entropy, mean absolute value (MAV), variance (VAR), modified mean absolute value type 1 (MAV1), waveform length (WL), simple square integral (SSI), Tsallis entropy, integrated EMG (IEMG) are extracted from the sub-bands. Extracted features are fed into an extreme learning machine (ELM) classifier with sigmoid activation function. ResultsComprehensive experiments are conducted on the input sEMG signals and the accuracy, sensitivity and specificity scores are used for performance measurement. Experiments showed that among all sub-bands, the seventh sub-band provided the best performance where the recorded accuracy, sensitivity and specificity values were 99.36%, 99.36% and 99.93%, respectively. The comparison results showed best efficiency of proposed method as compared to other methods on the same dataset. ConclusionThis paper investigates the usage of the FAWT and ELM on sEMG signal classification. The results show that the proposed method is quite efficient in classification of the sEMG signals. It is also observed that the seventh sub-band of the FAWT provides the best discrimination property. In the future works, recent wavelet transform methods will be used for improving the classification performance.

Event detection method comparison for distributed acoustic sensors using φ-OTDR

Abstract We investigate two kinds of event detection methods, support vector machine (SVM) and extreme learning machine (ELM), for distributed acoustic sensors (DAS) using phase-sensitive optical time domain reflectometer (φ-OTDR). We propose a heterodyne coherent φ-OTDR platform using an electric mixer, select spectrogram image feature (SIF) as a feature vector, and make comparison of the performance of 4 event classifiers. Theoretical and experimental analysis point out that an ELM classifier with RBF kernel is optimal in the detection of 5 sound events that has high identification accuracy, less training time, less testing time and strong noise immunity, which promises to be the first choice of event detection method in φ-OTDR systems in industrial applications.

Optimized Ensemble Extreme Learning Machine for Classification of Electrical Insulators Conditions

The classification of distinct problems of insulators in the distribution networks is a task that requires operator's experience. The applications of techniques to automate the inspection of electrical systems with the objective of detecting faults in insulators have shown to be reasonable alternatives to improve reliability in power grid. In this paper, based on the development of an experimental setup, signals are acquired considering three distinct faults in insulators. In this case, 13.8 kV (rms) is applied in drilled, contaminated, and good insulators considering an ultrasound detector connected to a computer. In the sequence, a multiclass classification method is proposed considering the ensemble of classifiers. The method considers the association of five distinct techniques, Bottom-Up segmentation, wavelet energy coefficient, principal component analysis, and particle swarm optimization associated with ensemble extreme learning machine (EN-ELM). Named optimized ensemble extreme learning machine, the present approach outperforms the original EN-ELM method. Finally, results show significant increase in robustness and faster training procedure when compared to classical approaches.

Annotating semantic tags of locations in location-based social networks

Location-based social networks (LBSNs) have become popular platforms that allow users to share their check-in activities with friends. Annotating semantic tags of locations, as one of the hottest research topics in LBSNs, has attracted considerable attention. Semantic annotation requires sufficient location features to train classifiers. Based on the analysis of LBSN data, we find that users’ check-in activities have similarities that can promote the extraction of location features and improve the accuracy of semantic annotation. However, the existing studies ignored the use of the similarities of users’ check-in activities for extracting suitable location features. Therefore, in this paper, a new location feature, called the similar user pattern (SUP), is first extracted by capturing the similarities among the different users’ check-in activities. Second, annotating semantic tags of locations is treated as a multi-label classification problem. Thus, multi-label semantic annotation with an extreme learning machine (ELM) is proposed, called MSA-ELM. The MSA-ELM algorithm trains a binary ELM classifier for each tag in the tag space to support multi-label classification. Finally, a series of experiments are conducted to demonstrate both the accuracy and efficiency of annotating semantic tags of locations. The experimental results show that the SUP is a proper location feature, and the MSA-ELM algorithm has a good performance for multi-label semantic annotation.

Daily reference evapotranspiration prediction based on hybridized extreme learning machine model with bio-inspired optimization algorithms: Application in contrasting climates of China

Reliable and accurate prediction of reference evapotranspiration (ETo) is a precondition for the efficient management and planning of agricultural water resources as well as the optimal design of irrigation scheduling. This study evaluated the performances of four bio-inspired algorithm optimized extreme learning machine (ELM) models, i.e. ELM with genetic algorithm (ELM-GA), ELM with ant colony optimization (ELM-ACO), ELM with cuckoo search algorithm (CSA) and ELM with flower pollination algorithm (ELM-FPA), for predicting daily ETo across China by using a five-fold cross-validation approach. These models were further compared with the classical ELM model parameterized by the grid search method to demonstrate their capability and efficiency. Daily maximum and minimum ambient temperatures, wind speed, relative humidity and global solar radiation data during 2001–2015 collected from eight meteorological stations in contrasting climates of China were utilized to train, validate and test the models. The results showed that ETo values predicted by all ELM models agreed well with the corresponding FAO-56 Penman–Monteith values, with R2, RMSE, NRMSE and MAE ranging 0.9766–0.9967, 0.0896–0.2883 mm day−1, 3.2910%-11.7653% and 0.0708–0.1998 mm day−1, respectively. The ELM-FPA model (R2 = 0.9930, RMSE = 0.1589 mm day−1, NRMSE = 5.5406% and MAE = 0.1188 mm day−1) slightly outperformed the ELM-CSA model (R2 = 0.9922, RMSE = 0.1619 mm day−1, NRMSE = 5.6864% and MAE = 0.1200 mm day−1) during testing, both of which were superior to the ELM-ACO (R2 = 0.9912, RMSE = 0.1730 mm day−1, NRMSE = 6.0816% and MAE = 0.1254 mm day−1) and ELM-GA (R2 = 0.9889, RMSE = 0.1895 mm day−1, NRMSE = 6.7197% and MAE = 0.1310 mm day−1) models, followed by the standalone ELM model (R2 = 0.9856, RMSE = 0.2104 mm day−1, NRMSE = 7.1693% and MAE = 0.1408 mm day−1). The four hybrid ELM models exhibited higher improvements in daily ETo prediction in the temperate monsoon and (sub)tropical monsoon climates (with average decrease in RMSE of 14.0%, 25.1%, 31.4% and 33.1%, respectively), compared with those in the temperate continental and mountain plateau climates (with average decrease in RMSE of 5.2%, 9.8%, 12.9% and 14.1%, respectively). The results advocated the capability of bio-inspired optimization algorithms, especially the FPA and CSA algorithms, for improving the performance of the conventional ELM model in daily ETo prediction in contrasting climates of China.

Automatic building change image quality assessment in high resolution remote sensing based on deep learning

The multi-temporal high-resolution remote sensing (HRRS) images are usually acquired at different imaging angles, with serious noise interferences and obvious building shadows, so that detecting the changes of urban buildings is a problem. In order to address this challenge, a deep learning-based algorithm called ABCDHIDL is proposed to automatically detect the building changes from multi-temporal HRRS images. Firstly, an automatic selection method of labeled samples of building changes based on morphology (ASLSBCM) is proposed. Secondly, a deep learning model (DBN-ELM) for building changes detection based on deep belief network (DBN) and extreme learning machine (ELM) is proposed. A convolution operation is employed to extract the spectral, texture and spatial features and generate a combined low-level features vector for each pixel in the multi-temporal HRRS images. The unlabeled samples are introduced to pre-train the DBN, and the parameters of DBN-ELM are globally optimized by jointly using the ELM classifier and the labeled samples are offered by ASLSBCM to further improve the detection accuracy. In order to evaluate the performance of ABCDHIDL, four groups of double-temporal WorldView2 HRRS images in four different experimental regions are selected respectively as the test datasets, and five other representative methods are used and compared with ABCDHIDL in the experiments of buildings change detection. The results show that ABCDHIDL has higher accuracy and automation level than the other five methods despite its relatively higher time consumption.