Multiview-Based Random Rotation Ensemble Pruning for Hyperspectral Image Classification

We employed the discrete wavelet transform to refl ectance spectra obtained from hyperspectral data to improve estimation of LAI in temperate forests. We estimated LAI for 32 plots across a range of forest types in Wisconsin using hemispherical photography. Plot spectra were extracted from AVIRIS data and transformed into wavelet features using the Haar wavelet. Separately, subsets of spectral bands and the Haar features selected by a genetic algorithm were used as independent variables in linear regressions. Models using wavelet coeffi cients explained the most variance for both broadleaf plots (R 2 = 0.90 for wavelet features versus R 2 = 0.80 for spectral bands) and all plots independent of forest type (R 2 = 0.79 for wavelet features vs. R 2 = 0.58 for spectral bands). The forest-type specifi c models were better than the models using all plots combined. Overall, wavelet features appear superior to band refl ectances alone for estimating temperate forest LAI using hyperspectral data.

Extreme Learning machine, ELM emerged as an efficient fast learning classifier for real valued classification problems. ELM suffers from the problem of instability due to random initialization of weights between the input and the hidden layer. Voting Based Extreme Learning Machine, VELM uses majority voting based ensembling technique to reduce this instability problem and improve the performance of ELM. VELM gives better performance at the cost of increased computational and memory requirement. This paper orders the component classifiers of the ensemble as per their importance using three different metrics: G-mean, Entropy and Margin based ordering. The pruned ensemble is constructed by selecting a threshold number of component classifiers as per their order of importance. The main aim of this work is to find which of these metric is more efficient for pruning VELM. This work presents an exhaustive analysis of these ordering metrics for varying sizes of pruned ensemble.

Gender recognition for interactive functions becomes essential topic in terms of service robotics applications. Ensemble learning which combines multiple classifiers prediction is now an active area of research in Machine Learning and Pattern Recognition. We propose an ensemble learning to facilitate gender classification. The features which we use are raw data (image pixels as input), Local binary pattern (LBP), Local derivative pattern (LDP), Gabor, and Weber local descriptors (WLD). We carry out comparative experimental studies of various gender recognition schemes, including Eigenfaces, any individual classifiers, Rotation Forest, and Adaboost. Specifically, not only individual classifiers but also ensemble classifiers are based on Support Vector Machine, namely using SVM as component classifier in ensemble learning. Furthermore, we evaluate the effect of image size on classification rate. In conclusion, we find that the best classification rate is achieved with Discrete Adaboost with SVM as component classifier using the aforementioned features. Another finding is that the classification rates will increase when face image size increases.

Designing adaptive classifiers for an evolving data stream is a challenging task due to the data size and its dynamically changing nature. Combining individual classifiers in an online setting, the ensemble approach, is a well-known solution. It is possible that a subset of classifiers in the ensemble outperforms others in a time-varying fashion. However, optimum weight assignment for component classifiers is a problem, which is not yet fully addressed in online evolving environments. We propose a novel data stream ensemble classifier, called Geometrically Optimum and Online-Weighted Ensemble (GOOWE), which assigns optimum weights to the component classifiers using a sliding window containing the most recent data instances. We map vote scores of individual classifiers and true class labels into a spatial environment. Based on the Euclidean distance between vote scores and ideal-points, and using the linear least squares (LSQ) solution, we present a novel, dynamic, and online weighting approach. While LSQ is used for batch mode ensemble classifiers, it is the first time that we adapt and use it for online environments by providing a spatial modeling of online ensembles. In order to show the robustness of the proposed algorithm, we use real-world datasets and synthetic data generators using the Massive Online Analysis (MOA) libraries. First, we analyze the impact of our weighting system on prediction accuracy through two scenarios. Second, we compare GOOWE with eight state-of-the-art ensemble classifiers in a comprehensive experimental environment. Our experiments show that GOOWE provides improved reactions to different types of concept drift compared to our baselines. The statistical tests indicate a significant improvement in accuracy, with conservative time and memory requirements.

Multiple classifier system exhibits strong classification capacity compared with single classifiers, but they require significant computational resources. Selective ensemble system aims to attain equivalent or better classification accuracy with fewer classifiers. However, current methods fail to identify precise solutions for constructing an ensemble classifier. In this study, we propose an ensemble classifier design technique based on the perturbation binary salp swarm algorithm (ECDPB). Considering that extreme learning machines (ELMs) have rapid learning rates and good generalization ability, they can serve as the basic classifier for creating multiple candidates while using fewer computational resources. Meanwhile, we introduce a combined diversity measure by taking the complementarity and accuracy of ELMs into account; it is used to identify the ELMs that have good diversity and low error. In addition, we propose an ECDPB with powerful optimizing ability; it is employed to find the optimal subset of ELMs. The selected ELMs can then be used to form an ensemble classifier. Experiments on 10 benchmark datasets have been conducted, and the results demonstrate that the proposed ECDPB delivers superior classification capacity when compared with alternative methods.

Tree species mapping in forest areas is an important topic in forest inventory. In recent years, several studies have been carried out using different types of hyperspectral sensors under various forest conditions. The aim of this work was to evaluate the potential of two high spectral and spatial resolution hyperspectral sensors (HySpex-VNIR 1600 and HySpex-SWIR 320i), operating at different wavelengths, for tree species classification of boreal forests. To address this objective, many experiments were carried out, taking into consideration: 1) three classifiers (support vector machines (SVM), random forest (RF), and Gaussian maximum likelihood); 2) two spatial resolutions (1.5 m and 0.4 m pixel sizes); 3) two subsets of spectral bands (all and a selection); and 4) two spatial levels (pixel and tree levels). The study area is characterized by the presence of four classes 1) Norway spruce, 2) Scots pine, together with 3) scattered Birch and 4) other broadleaves. Our results showed that: 1) the HySpex VNIR 1600 sensor is effective in boreal tree species classification with kappa accuracies over 0.8 (with Pine and Spruce reaching producer's accuracies higher than 95%); 2) the role of the HySpex-SWIR 320i is limited, and its bands alone are able to properly separate only Pine and Spruce species; 3) the spatial resolution has a strong effect on the classification accuracy (an overall decrease of more than 20% between 0.4 m and 1.5 m spatial resolution); and 4) there is no significant difference between SVM or RF classifiers.

Currently, most text classifiers apply machine learning methods, while ignore traditional methods based on classification rules. In this paper, we propose a strong covering algorithm (called SCA) for generating strong classification rules and view the rules-based classifier as a component classifier in the ensemble text classifier. SCA extracts noun phrase to index document based-on our proposed Exhaustive Noun-Phrase Extraction Algorithm. Experimental results show that the ensemble classifier integrating the strong rules achieves an approximately 8% improvement as compared to bi-gram classifier and 15% improvement as compared to the single rule-based classifier.

Ensemble classifiers are a promising approach for data stream classification. Though, diversity influences the performance of ensemble classifiers, current studies do not take advantage of relations between component classifiers to improve their performance. This paper addresses this issue by proposing a new kind of ensemble learner for data stream classification, which explicitly defines relations between component classifiers. These relations are then used in various ways, e.g., to combine the decisions of component models. The hypothesis is that an ensemble learner can yield accurate predictions in a streaming environment based on a structural analysis of a weighted network of its component models. Implications, limitations and benefits of this assumption, are discussed. A formal description of a network-based ensemble for data streams is presented, and an algorithm that implements it, named Network of Experts (NetEx). Empirical experiments show that NetEx’s accuracy and processing time are competitive with state-of-the-art ensembles.

This paper presents a novel algorithm named diverse AdaBoostSVM tracking (DABSVT) for target tracking in infrared imagery. The tracker trains a support vector machine (SVM) classifier per frame. All of the classifiers are combined into an ensemble classifier using AdaBoost. By proper parameter adjusting strategies, a set of effective SVM classifiers with moderate accuracy are obtained, and the dilemma problem between accuracy and diversity of AdaBoost is dealt with too. To cope with the changes in features of both foreground and background, the component classifier can be discarded or added at any time. The experiments performed on several sequences show the robustness of the proposed method

Dissimilarity based ensemble of extreme learning machine for gene expression data classification

For classification problem, extreme learning machine (ELM) can get better generalization performance at a much faster learning speed. Nevertheless, a single ELM is unstable in data classification. The Bagging-based ensemble classifier, i.e., Bagging-ELM has been studied popularly and proved to improve the performance of ELM significantly in terms of accuracy, however, it is inappropriate to deal with large-scale datasets due to the highly intensive computation. In this study, we propose a novel ELM ensemble classifier, namely b-ELM, which leverages the Bag of Little Bootstraps technique to obtain a scalable, efficient means of classification for large-scale data. Efficiency of classification is achieved as it only requires repeated training under consideration on quantities of data that can be much smaller than the original training data. Furthermore, b-ELM is suited to implementation on modern parallel and distributed computing platforms. The experimental results demonstrate that b-ELM can efficiently handle large-scale data with a good performance on prediction accuracy.

Currently, irrigation decisions in coffee cultivation primarily rely on empirical knowledge, resulting in inefficient practices. Combining real-time leaf water content (LWC) data can help improve the accuracy of the irrigation planning. Spectral remote sensing is a fast, reliable, and non-invasive method to detect vegetation moisture content. In this study, a model to estimate the LWC of Coffea arabica L. was built using hyperspectral reflectance of the canopy under various irrigation levels. For this purpose, common spectral indices, two-band spectral indices [ratio spectral index (RSI); difference spectral index (DSI); and normalized difference spectral index (NDSI)], and three-band spectral indices were constructed. Feature bands were extracted using the successive projections algorithm (SPA). Optimal spectral indices were extracted using the correlation coefficient method, and the feature wavebands and spectral indices were combined into five datasets. These datasets were split into modeling and validation datasets by sample set partitioning based on the joint x-y distance (SPXY) algorithm. A linear model [partial least squares regression (PLSR)] and three non-linear models [support vector machine (SVM); extreme learning machine (ELM); back propagation artificial neural network (BPANN)] were built to estimate LWC of Coffea arabica L. The results indicated that the non-linear models surpassed the linear model. The accuracy was the highest when the modeling was performed using the dataset combination 5. Among various modeling methods, the predictive performance of ELM was the best (modeling dataset: R2=0.745, RMSE=2.241%, RRMSE=3.482%; validation dataset: R2=0.721, RMSE=2.142%, RRMSE=3.364%). ELM outperformed PLSR, SVM, and BPANN in LWC retrieval. The obtained results indicated that the dataset built by the combined use of different methods was superior to those from a single data source in accuracy. This study provides a scientific basis for the quantitative diagnosis of coffee tree water status, with significant implications for optimizing field irrigation management. Keywords: spectral index, feature band, dataset, Coffea arabica L., leaf water content, machine learning DOI: 10.25165/j.ijabe.20251805.9067 Citation: Liu X G, Peng K L, Chen S M, Tuo Y F, Zhang S, Tan S, et al. Estimating the leaf water content of Coffea arabica L. based on hyperspectral reflectance and dataset construction. Int J Agric & Biol Eng, 2025; 18(5): 287–297.

Abstract— This paper introduces a comprehensive framework for automated disease detection using pupillometry data. Our approach establishes a robust pipeline that includes data preprocessing, feature extraction, and machine learning-based classification of patients based on their pupillary responses. We extract key features from both left and right pupil diameter measurements, such as maximum and minimum values, delta, channel height (CH), latency, and mean change velocity (MCV).To enhance classification accuracy, we train and evaluate multiple machine learning models, including Support Vector Machines (SVMs), an ensemble classifier, Extreme Learning Machines (ELM), Multi-Layer Perceptrons (MLPs), and Random Forests. Additionally, we propose a novel hybrid model that integrates the strengths of multiple algorithms, outperforming individual models in accuracy. Our experimental results highlight the effectiveness of this hybrid approach, demonstrating its potential for improving non-invasive and efficient disease diagnosis. This research contributes to advancements in clinical ophthalmology and neurology by leveraging pupillometry and machine learning for more precise and accessible diagnostic tools. Keywords: Pupillometry data, Channel height, Latency, Mean Change Velocity, Support Vector Machines, Extreme Learning Machines, Multi Layer Perceptrons, Random Forest.

Technology-sourcing mergers and acquisitions (M&As) has become an important instrument for enterprises to acquire new knowledge and build technological ability. To achieve technology innovation improvement, the choice of integration strategy is vitally important. The purpose of this paper is to explore the relationship between integration and innovation in technology-sourcing M&As. To verify our hypotheses, we conduct a comparative empirical study based on the data of overseas and domestic M&As of Chinese enterprises from 2000 to 2019. We find that for overseas M&As, the influence of integration degree on technology innovation depends on the resource characteristics of the two merging parties. In cases of high resource similarity and low resource complementarity, the degree of integration positively affects technology innovation, while in cases of low resource similarity and high resource complementarity, the degree of integration negatively affects technology innovation. In comparison, for domestic M&As, the degree of integration positively affects technology innovation.

In order to improve the classification of hyperspectral image(HSI), we propose a novel hyperspectral image classification method based on the comprehensive evaluation model of extreme learning machine(ELM) with the cumulative variation weights(CVW), referred to as ELM with the cumulative variation weights and comprehensive evaluation (CVW-CEELM). To be specific, the cumulative variation value is proposed as a new metric. The inefficient bands are eliminated by the cumulative variation quotient values based on the cumulative variation values. The cumulative variation weights based on the cumulative variation values are used to determine the contribution of each weak ELM classifier to the hyperspectral image classification algorithm. The remaining effective bands are divided by grouping strategy. In each group of the effective bands, the different numbers of bands are selected to reduce the dimension of the hyperspectral image dataset by the weighted random-selecting-based method. After dimensionality reduction, the spatial-spectral features of each pixel are extracted and multiple weak ELM classifiers are trained by the training samples. Then, the results of several weak classifiers are synthetically evaluated by the cumulative variation weights to get the final classification results. Experimental results on the typical hyperspectral image datasets illustrate that the proposed CVW-CEELM has few adjustable parameters to make the operation simple, and outperforms a variety of the image classification counterparts in terms of the calculation cost and classification accuracy.