Classical Benchmark Research Articles

A hybrid Prairie INFO fission naked algorithm with stagnation mechanism for the parametric estimation of solar photovoltaic systems

This paper presents a study to enhance the performance of a recently introduced naked mole-rat algorithm (NMRA), by local optima avoidance, and better exploration as well as exploitation properties. A new set of algorithms, namely Prairie dog optimization algorithm, INFO, and Fission fusion optimization algorithm (FuFiO) are included in the fundamental framework of NMRA to enhance the exploration operation. The proposed algorithm is a hybrid algorithm based on four algorithms: Prairie Dog, INFO, Fission Fusion and Naked mole-rat (PIFN) algorithm. Five new mutation operators/inertia weights are exploited to make the algorithm self-adaptive in nature. Apart from that, a new stagnation phase is added for local optima avoidance. The proposed algorithm is tested for variable population, dimension size, and efficient set of parameters is analysed to make the algorithm self-adaptive in nature. Friedman as well as Wilcoxon rank-sum tests are performed to determine the effectiveness of the PIFN algorithm. On the basis of a comparison of outcomes, the PIFN algorithm is more effective and robust than the other optimization techniques evaluated by prior researchers to address standard benchmark functions (classical benchmarks, CEC 2017, and CEC-2019) and complex engineering design challenges. Furthermore, the effectiveness as well as reliability of the PIFN algorithm is demonstrated by testing using various PV modules, namely the RTC France Solar Cell (SDM, and DDM), Photowatt-PWP201, STM6- 40/36, and STP6-120/36 module. The results obtained from the PIFN algorithm are compared with various MH algorithms reported in the existing literature. The PIFN algorithm achieved the lowest root-mean-square error value, for RTC France Solar Cell (SDM) is 7.72E−04, RTC France Solar Cell (DDM) is 7.59E−04, STP6-120/36 module is 1.44E−02, STM6-40/36 module is 1.723E−03, and Photowatt-PWP201 module is 2.06E−03, respectively. In order to enhance the accuracy of the obtained results of parameter estimation of solar photovoltaic systems, we integrated the Newton-Raphson approach with the PIFN algorithm. Experimental and statistical results further prove the significance of the PIFN algorithm with respect to other algorithms.

Research on Intelligent Optimization of Wellbore Trajectory in Complex Formation

Borehole trajectory optimization is a key issue in oil and gas drilling engineering. The traditional wellbore trajectory design method faces great challenges in optimizing the trajectory length and complexity, and it is difficult to meet the actual engineering requirements. In this paper, the three-stage wellbore trajectory optimization problem is studied, and a multi-objective optimization model including two objective functions of trajectory length and trajectory complexity is constructed. In this paper, an improved multi-objective particle swarm optimization algorithm is proposed, which combines the clustering strategy to improve the diversity of solutions, and enhances the local search ability and global convergence performance of the algorithm through the elite learning strategy. In order to verify the performance of the algorithm, comparative experiments were carried out using classical multi-objective benchmark functions. The results showed that the improved algorithm is superior to the traditional method in terms of diversity and convergence of solutions. Finally, the proposed algorithm was applied to the actual three-stage wellbore trajectory optimization problem. In summary, the research results of this paper provide theoretical support and engineering practice methods for wellbore trajectory optimization, and serve as an important reference for further improving the efficiency and quality of wellbore trajectory design.

Graph Transformations for Memory Peak Minimization by Scheduling

Many computing systems are constrained by a fixed amount of available shared memory. Modeling applications with task graphs or Synchronous DataFlow (SDF) graphs makes it possible to analyze and optimize their memory usage. The NP-complete problem studied here is to find a sequential schedule of dataflow graphs that minimizes their memory peak. We propose four task graph transformations that reduce the graph size while preserving the minimal memory peak. They are able to compress all Series-Parallel Directed Acyclic Graphs (SP-DAG) to a single node representing an optimal schedule (optimal in the sense that its memory peak is minimal). They also offer simple criteria to schedule optimally independent tasks and guided us to develop an optimal compositional scheduling analysis. These transformations are quite effective and compress a large class of graphs into a single node representing an optimal schedule. When this is not the case, we propose an optimized branch and bound algorithm to complete the analysis of the reduced graphs. This algorithm is able to find the optimal schedule of graphs comprising up to 100 tasks. Our approach also applies to SDF graphs after converting them to task graphs. However, since this conversion may produce huge graphs, we also describe a new suboptimal method, similar to Partial Expansion Graphs, to reduce the graph and schedule sizes. We evaluate our approach on classic benchmarks, on which we always outperform the state-of-the-art.

Optimization of manipulator inverse kinematics using improved particle swarm algorithm for enhanced manufacturing efficiency

In the context of digital manufacturing and intelligent manufacturing systems, optimizing the inverse kinematics (IK) of manipulators is crucial for improving manufacturing efficiency and productivity. The authors propose an improved particle swarm optimizer (IPSO) with adaptive inertia weight and asynchronous learning factors to enhance the optimization performance. The proposed algorithm introduces a novel mutation strategy where selected initial particles undergo mutation to generate new particles, thereby increasing swarm diversity. The best-performing particles after mutation replace the initial ones to improve global optimization capability. The effectiveness of the IPSO algorithm is evaluated through thirteen classical benchmark functions and compared with genetic algorithms (GA), standard particle swarm optimization (PSO), linear decreasing weight PSO (LDWPSO), and biogeography-based PSO (BLPSO), quantum-behaved particle swarm optimization (QPSO), Levy Flight quantum-behaved particle swarm optimization (LQPSO), Gaussian quantum-behaved particle swarm optimization approaches (GQPSO). Experimental mean results demonstrate that the IPSO algorithm outperforms these algorithms. The IPSO algorithm and the other algorithms subsequently were employed to estimate the performance for tackling the IK problem of a manipulator and the comparison results demonstrate the superior performance of IPSO algorithm in terms of accuracy and convergence rate. Finally, the practical efficacy of the IPSO algorithm was validated through experimental trials conducted on a custom-designed six-degree-of-freedom (6-DOF) collaborative robotic manipulator platform, which demonstrated the viability of IPSO algorithm in real-world manufacturing applications. The findings suggest that the proposed IPSO algorithm offers a promising solution for enhancing manipulator control efficiency in digital manufacturing systems. The IPSO code is available at: https://10.6084/m9.figshare.27753549 .

SEMG Signal-Based Human Lower-Limb Intention Recognition Algorithm Using Improved Extreme Learning Machine

To address the difficulty in identifying human lower-limb movement intentions, low accuracy of classification models, and weak generalization ability, this study proposes a motion intention recognition method that combines an improved pelican optimization algorithm (IPOA) and a hybrid kernel extreme learning machine (HKELM). First, we collect the surface electromyography (sEMG) signals of subjects in six motion modes and perform feature parameter extraction under non-ideal conditions. On this basis, we establish a dataset of the relationships between the feature parameters and gait movements. Second, we build a motion intention classification model based on relational data using the HKELM to solve the problems of low modeling accuracy and weak generalizability. Third, the IPOA is used to optimize the parameters related to the HKELM, and a differential evolution algorithm is introduced to improve the population quality and prevent the algorithm from falling into a local optimal solution. The experimental results show that the IPOA exhibits better optimization accuracy and convergence speed for four classical benchmark functions. Its average classification accuracy, average classification recall, and average F-value are 94.45%, 94.47%, and 94.46%, respectively, which are significantly higher than those of other intention recognition algorithms. Therefore, the proposed method has high classification accuracy and generalization performance.

Natural Language Processing for Dialects of a Language: A Survey

State-of-the-art natural language processing (NLP) models are trained on massive training corpora, and report a superlative performance on evaluation datasets. This survey delves into an important attribute of these datasets: the dialect of a language. Motivated by the performance degradation of NLP models for dialectal datasets and its implications for the equity of language technologies, we survey past research in NLP for dialects in terms of datasets, and approaches. We describe a wide range of NLP tasks in terms of two categories: natural language understanding (NLU) (for tasks such as dialect classification, sentiment analysis, parsing, and NLU benchmarks) and natural language generation (NLG) (for summarisation, machine translation, and dialogue systems). The survey is also broad in its coverage of languages which include English, Arabic, German, among others. We observe that past work in NLP concerning dialects goes deeper than mere dialect classification, and extends to several NLU and NLG tasks. For these tasks, we describe classical machine learning using statistical models, along with the recent deep learning-based approaches based on pre-trained language models. We expect that this survey will be useful to NLP researchers interested in building equitable language technologies by rethinking LLM benchmarks and model architectures.

The importance of sampling the dynamical modes: Reevaluating benchmarks for invariant and equivariant features of machine learning potentials for simulation of free energy landscapes.

Machine learning interatomic potentials (MLIPs) are rapidly gaining interest for molecular modeling, as they provide a balance between quantum-mechanical level descriptions of atomic interactions and reasonable computational efficiency. However, questions remain regarding the stability of simulations using these potentials, as well as the extent to which the learned potential energy function can be extrapolated safely. Past studies have encountered challenges when MLIPs are applied to classical benchmark systems. In this work, we show that some of these challenges are related to the characteristics of the training datasets, particularly the inefficient exploration of the dynamical modes and the inclusion of rigid constraints. We demonstrate that long stability in simulations with MLIPs can be achieved by generating unconstrained datasets using unbiased classical simulations, provided that the important dynamical modes are correctly sampled. In addition, we emphasize that in order to achieve precise energy predictions, it is important to resort to enhanced sampling techniques for dataset generation, and we demonstrate that safe extrapolation of MLIPs depends on judicious choices related to the system's underlying free energy landscape and the symmetry features embedded within the machine learning models.

Enhanced Nutcracker Optimization Algorithm with Hyperbolic Sine-Cosine Improvement for UAV Path Planning.

Three-dimensional (3D) path planning is a crucial technology for ensuring the efficient and safe flight of UAVs in complex environments. Traditional path planning algorithms often find it challenging to navigate complex obstacle environments, making it challenging to quickly identify the optimal path. To address these challenges, this paper introduces a Nutcracker Optimizer integrated with Hyperbolic Sine-Cosine (ISCHNOA). First, the exploitation process of the sinh cosh optimizer is incorporated into the foraging strategy to enhance the efficiency of nutcracker in locating high-quality food sources within the search area. Secondly, a nonlinear function is designed to improve the algorithm's convergence speed. Finally, a sinh cosh optimizer that incorporates historical positions and dynamic factors is introduced to enhance the influence of the optimal position on the search process, thereby improving the accuracy of the nutcracker in retrieving stored food. In this paper, the performance of the ISCHNOA algorithm is tested using 14 classical benchmark test functions as well as the CEC2014 and CEC2020 suites and applied to UAV path planning models. The experimental results demonstrate that the ISCHNOA algorithm outperforms the other algorithms across the three test suites, with the total cost of the planned UAV paths being lower.

Comparative analysis of accuracy and computational complexity across 21 swarm intelligence algorithms

Nonlinear, complex optimization problems are prevalent in many scientific and engineering fields. Traditional algorithms often struggle with these problems due to their high dimensionality and intricate nature, making them time-consuming. Many researchers have proposed new metaheuristic algorithms inspired by biological behaviors in nature, which comparatively show higher performance and accuracy than traditional optimization algorithms. Nature-inspired algorithms, particularly those based on swarm intelligence, offer adaptable and efficient solutions to these challenges. In recent years, swarm intelligence algorithms have made significant advancements. Classical and CEC benchmark suits are immersively useful for studying the performance of optimization algorithms. According to our literature survey, we identified that many algorithms were evaluated based on accuracy. Currently, swarm intelligence algorithms are used in many applications, and efficiency and computational complexity need to be evaluated. A broad-level study of the computational complexity and accuracy of popular swarm intelligence algorithms has not been done recently. Therefore this study we comprehensively evaluate and compare 21 bio-inspired swarm intelligence algorithms on eight non-separable unimodal, eight separable unimodal, five non-separable multimodal, seven separable multimodal functions, and two CEC 2018 many objective functions. We study the structure and mathematical model of the selected algorithms. Then we categorized selected algorithms into six different behavioral groups. We calculated the root mean square error between expected and actual values. Then we performed an RMSE cross-validation statistical test to understand how accurately an algorithm resolves an average problem. We found that Artificial Lizard Search Optimization (ALSO) is the most prominent algorithm in accuracy and efficiency. Besides that, Cat Swarm Optimization (CSO), Squirrel Search Algorithm (SSA), and Chimp Optimization Algorithm (CHOA-B) are also considered more universal algorithms. The Squirrel Search Algorithm (SSA) is ALSO’s second-best algorithm in time complexity. Wasp Swarm Algorithm (WSO), and Bat-Inspired Algorithm (BA) presented the lowest time complexity. Finally, several important issues and research directions are discussed.

Applications of the Circle-Inspired Optimization Algorithm: Comparison between MatLab and Python versions

This paper presents applications of the Circle Inspired Optimization Algorithm (CIOA), a modern optimization algorithm developed by the authors, in optimization problems from different areas of study implemented in MatLab and Python, with the main objective of making a comparison between versions of each computer language. Different sets of optimization problems were selected, involving classical benchmark functions, traditional truss structural optimization problems, and real-world optimization problems obtained in the 'CEC2020 Real-World One-Objective Constrained Optimization Competition'. Each optimization problem was solved multiple times on the same computer in MatLab and Python, to make comparisons between the accuracy and robustness of each version of the algorithm, evaluating the best solution and the mean and standard deviation between several solutions. Furthermore, a comparison of computational time was also performed. The results obtained attest to the efficiency of CIOA in both computational languages. In terms of accuracy and robustness, it was not possible to clearly identify a better version (MatLab or Python) for CIOA, as the differences obtained in the results were, in most cases, very small, and may be caused by the random characteristics of the algorithm. In the comparison of computational time, the MatLab version performed better in most optimization problems. This fact may be due to the efficiency of the specific libraries used and the programming experience of the authors in each computer language.

Automatic Detection and Classification of Aurora in THEMIS All‐Sky Images

AbstractWe report a novel machine‐learning algorithm for automatically detecting and classifying aurora in all–sky images (ASI) that is largely trained without requiring ground–truth labels. By including a small number of labeled images, we are able to automatically label all of the approximately 700 million images in the Time History of Events and Macroscale Interactions during Substorms (THEMIS) ASI data set from 2008 to 2022. We use a two–stage approach. In the first stage, we adapt the Simple framework for Contrastive Learning of Representations (SimCLR) algorithm to learn latent representations of THEMIS all–sky images. We then finetune a classifier network on the latent representations our model learns of the manually labeled Oslo aurora THEMIS (OATH) data set. We demonstrate that this two–stage approach achieves excellent classification results on data for which there is no current ML classification benchmark. The outcome of this work will facilitate efficient information retrieval for researchers interested in specific categories of aurora and will enable large scale statistical studies and machine learning analyses of THEMIS all–sky images that have not previously been possible. To demonstrate possible ways to utilize this database, we performed a statistical analysis of the occurrence rates of auroral labels with respect to solar wind parameters, interplanetary magnetic field vector, and geomagnetic indices. We further investigate the occurrence rates of auroral phenomena in the annotated data set and their geoeffectiveness by utilizing the co–located THEMIS ground magnetometer data set.

Bypassing Stationary Points in Training Deep Learning Models.

Gradient-descent-based optimizers are prone to slowdowns in training deep learning models, as stationary points are ubiquitous in the loss landscape of most neural networks. We present an intuitive concept of bypassing the stationary points and realize the concept into a novel method designed to actively rescue optimizers from slowdowns encountered in neural network training. The method, bypass pipeline, revitalizes the optimizer by extending the model space and later contracts the model back to its original space with function-preserving algebraic constraints. We implement the method into the bypass algorithm, verify that the algorithm shows theoretically expected behaviors of bypassing, and demonstrate its empirical benefit in regression and classification benchmarks. Bypass algorithm is highly practical, as it is computationally efficient and compatible with other improvements of first-order optimizers. In addition, bypassing for neural networks leads to new theoretical research such as model-specific bypassing and neural architecture search (NAS).

ELODI: Ensemble Logit Difference Inhibition for Positive-Congruent Training.

Negative flips are errors introduced in a classification system when a legacy model is updated. Existing methods to reduce the negative flip rate (NFR) either do so at the expense of overall accuracy by forcing a new model to imitate the old models, or use ensembles, which multiply inference cost prohibitively. We analyze the role of ensembles in reducing NFR and observe that they remove negative flips that are typically not close to the decision boundary, but often exhibit large deviations in the distance among their logits. Based on the observation, we present a method, called Ensemble Logit Difference Inhibition (ELODI), to train a classification system that achieves paragon performance in both error rate and NFR, at the inference cost of a single model. The method distills a homogeneous ensemble to a single student model which is used to update the classification system. ELODI also introduces a generalized distillation objective, Logit Difference Inhibition (LDI), which only penalizes the logit difference of a subset of classes with the highest logit values. On multiple image classification benchmarks, model updates with ELODI demonstrate superior accuracy retention and NFR reduction.

Tuning Vision-Language Models With Multiple Prototypes Clustering.

Benefiting from advances in large-scale pre-training, foundation models, have demonstrated remarkable capability in the fields of natural language processing, computer vision, among others. However, to achieve expert-level performance in specific applications, such models often need to be fine-tuned with domain-specific knowledge. In this paper, we focus on enabling vision-language models to unleash more potential for visual understanding tasks under few-shot tuning. Specifically, we propose a novel adapter, dubbed as lusterAdapter, which is based on trainable multiple prototypes clustering algorithm, for tuning the CLIP model. It can not only alleviate the concern of catastrophic forgetting of foundation models by introducing anchors to inherit common knowledge, but also improve the utilization efficiency of few annotated samples via bringing in clustering and domain priors, thereby improving the performance of few-shot tuning. We have conducted extensive experiments on 11 common classification benchmarks. The results show our method significantly surpasses the original CLIP and achieves state-of-the-art (SOTA) performance under all benchmarks and settings. For example, under the 16-shot setting, our method exhibits a remarkable improvement over the original CLIP by 19.6%, and also surpasses TIP-Adapter and GraphAdapter by 2.7% and 2.2%, respectively, in terms of average accuracy across the 11 benchmarks.

A variationally-consistent hybrid equilibrium element formulation for linear poroelasticity

A poroelastic medium is defined as a continuous system in which the mechanical response arises from the interaction between a deformable elastic solid skeleton and a pressurised fluid, fully saturating the interconnected porous network. The coupled theory of Poromechanics is effectively employed to solve a broad class of problems spanning various fields, ranging from its original application in Geomechanics to addressing contemporary challenges in tissue Biomechanics. Besides the seminal theory introduced by Biot, several variational multi-field principles have been proposed, leading to various finite element formulations. Finite Element approaches are mostly based on a two-field discretisation involving the coupled solid displacement and interstitial fluid pressure as primary variables. However, this choice of the two-primary variables leads to a saddle-point problem, posing a challenge in achieving accurate solutions and thus making the use of stabilisation methods rather mandatory. This study proposes a novel variationally consistent Hybrid Equilibrium Element formulation, based on the complementary strain energy function. A two-field minimisation principle is formulated, with total Cauchy stress and interstitial fluid pressure serving as primary variables. The proposed variational principle provides a variationally consistent framework for designing inherently self-equilibrated Hybrid Equilibrium Elements, where the solid displacement field emerges as a Lagrangian variable to enforce a co-diffusivity constraint at element sides. The fulfilment of the inf-sup condition, typically challenging in developing stable and computationally efficient finite element formulations, is no longer required, a minimisation principle having been adopted, rather than the classical saddle-point principle. The proposed Hybrid Equilibrium Element approach offers a statically admissible solution even in the proximity of singularities or high space gradient stress distribution, providing a more comprehensive assessment of stress distributions and concentrations. The convexity of the new variational principle has been established, ensuring solution uniqueness. The formulation is based on a 2-D triangular Hybrid Equilibrium Element employed to solve plane strain problems. The element has been implemented in the Open source Finite Element Analysis Program (FEAP) and its performance evaluated against classical benchmark problems for poroelasticity, with the emphasis on stress accuracy. The numerical results obtained for the quasi-static analysis of poroelastic systems show the advantages of the proposed approach, in which accurate stress distributions are displayed, even adopting quite coarse meshes. Remarkably, in the proposed approach no stabilisation technique is required.

A novel deep reinforcement learning‐based algorithm for multi‐objective energy‐efficient flow‐shop scheduling

AbstractA novel algorithm combining bidirectional recurrent neural networks (BiRNNs) with temporal difference is proposed for multi‐objective energy‐efficient non‐permutation flow‐shop scheduling problem (NFSP). The objectives of this problem involve minimising both the makespan and total energy consumption. To begin, a mathematical model is formulated to represent the energy‐efficient NFSP. Subsequently, the NFSP is transformed into a Markov decision process, where an action space comprising 28 scheduling rules is constructed. Considering the global and local features of NFSP, a set of 15 state features is extracted. Different reward functions are then defined to correspond to the specific objectives. Furthermore, the state features of NFSP are extracted using a multi‐layer perceptron model based on BiRNNs. By utilising the TD(λ) algorithm to calculate the state value function, various policies are generated. In order to evaluate the proposed algorithm, a new test set for the energy‐efficient NFSP is constructed, building upon classic benchmark problems. Finally, comparison experiments are conducted to demonstrate the effectiveness and efficiency of the proposed algorithm.

Modelling unstable crack propagation in concrete by finite element method with continuous nodal stress

Cracks in concrete will propagate unstably due to excessive shear stress, interactions at the rock-concrete interface, or sudden energy release, significantly compromising structural bearing capacity. To investigate this phenomenon, a nonlinear numerical method for modelling mixed-mode I-II crack propagation has been developed. This approach integrates dynamic equilibrium with a fictitious crack model and an initiation fracture toughness criterion. Key innovations include the incorporation of a finite element method with continuous nodal stress to enhance calculation accuracy, the utilization of kinetic energy to compensate for abrupt losses of strain energy during crack propagation, and the consideration of the deformation of the distributive beam and its interactions with specimens and supports. It was validated by modelling classic benchmarks for the dynamic initiation and propagation of brittle materials under mode I and mixed-mode I-II loading, and applied to analyse unstable crack propagation in concrete for four-point shear (FPS) beam specimens under various ratios of mode I and II stress intensity factors (KI/KII = 0 to 5.32) and loading rates (2 × 10-7 m/s to 2 × 10-3 m/s). Results indicated that the method effectively captures unstable crack propagation, with load-crack mouth shear displacement (CMSD) curves and crack propagation trajectories closely matching the experimental data. Furthermore, it was observed that when the elastic strain energy within the concrete beam exceeds the residual energy stored in the extended cracks, the crack transitions from stability to instability; conversely, if the elastic strain energy is less than or approximately equal to the residual energy, the crack decelerates and returns to stability. Additionally, parametric analyses reveal that lower distributive beam stiffness, shorter preset crack lengths, reduced concrete fracture energy, and a less robust cohesive force–displacement curve increase the likelihood of unstable crack propagation in concrete.

Scattering wave packets of hadrons in gauge theories: Preparation on a quantum computer

Quantum simulation holds promise of enabling a complete description of high-energy scattering processes rooted in gauge theories of the Standard Model. A first step in such simulations is preparation of interacting hadronic wave packets. To create the wave packets, one typically resorts to adiabatic evolution to bridge between wave packets in the free theory and those in the interacting theory, rendering the simulation resource intensive. In this work, we construct a wave-packet creation operator directly in the interacting theory to circumvent adiabatic evolution, taking advantage of resource-efficient schemes for ground-state preparation, such as variational quantum eigensolvers. By means of an ansatz for bound mesonic excitations in confining gauge theories, which is subsequently optimized using classical or quantum methods, we show that interacting mesonic wave packets can be created efficiently and accurately using digital quantum algorithms that we develop. Specifically, we obtain high-fidelity mesonic wave packets in the Z2 and U(1) lattice gauge theories coupled to fermionic matter in 1+1 dimensions. Our method is applicable to both perturbative and non-perturbative regimes of couplings. The wave-packet creation circuit for the case of the Z2 lattice gauge theory is built and implemented on the Quantinuum H1-1 trapped-ion quantum computer using 13 qubits and up to 308 entangling gates. The fidelities agree well with classical benchmark calculations after employing a simple symmetry-based noise-mitigation technique. This work serves as a step toward quantum computing scattering processes in quantum chromodynamics.

Euler Kernel Mapping for Hyperspectral Image Clustering via Self-Paced Learning

Clustering, as a classical unsupervised artificial intelligence technology, is commonly employed for hyperspectral image clustering tasks. However, most existing clustering methods designed for remote sensing tasks aim to solve a non-convex objective function, which can be optimized iteratively, beginning with random initializations. Consequently, during the learning phase of the clustering model, it may easily fall into bad local optimal solutions and finally hurt the clustering performance. Additionally, prevailing approaches often exhibit limitations in capturing the intricate structures inherent in hyperspectral images and are very sensitive to noise and outliers that widely exist in remote sensing data. To address these issues, we proposed a novel Euler kernel mapping for hyperspectral image clustering via self-paced learning (EKM-SPL). EKM-SPL first employs self-paced learning to learn the clustering model in a meaningful order by progressing samples from easy to complex, which can help to remove bad local optimal solutions. Secondly, a probabilistic soft weighting scheme is employed to measure complexity across the data sample, which makes the optimization process more reasonable. Thirdly, in order to more accurately characterize the intricate structure of hyperspectral images, Euler kernel mapping is used to convert the original data into a reproduced kernel Hilbert space, where the nonlinearly inseparable clusters may become linearly separable. Moreover, we innovatively integrate the coordinate descent technique into the optimization algorithm to circumvent the computational inefficiencies and information loss typically associated with conventional kernel methods. Extensive experiments conducted on classic benchmark hyperspectral image datasets illustrate the effectiveness and superiority of our proposed model.

Poverty Classification in Indonesia Using BiGRU, BPNN, and Stacking AdaBoost Frameworks

This research addresses the persistent global challenge of poverty, with a specific focus on Indonesia, a nation with a population exceeding 270 million. The primary objective is to enhance the precision and reliability of poverty classification using advanced machine learning technologies. We employed a combination of Bidirectional Gated Recurrent Unit (BiGRU), Backpropagation Neural Network (BPNN), and stacking techniques with AdaBoost to develop an innovative classification model. The methodology involved training each technique separately and then integrating them into a stacked model to leverage their individual strengths. The results were promising, demonstrating a substantial improvement in model performance with precision, recall, and F1 scores reaching as high as 0.98, and an overall accuracy of 98.06%. The use of visual analytics, including pie charts and bar graphs, provided a comprehensive distribution analysis of poverty levels, confirming the balanced nature of the dataset. These findings underscore the critical role of machine learning in formulating effective policies for poverty alleviation and suggest that integrating multiple machine learning algorithm can significantly enhance decision-making processes. The novelty of this research lies in the successful application of a stacked machine learning model combining BiGRU, BPNN, and AdaBoost, which establishes a new benchmark for poverty classification in large-scale social datasets. This study not only contributes to the academic discourse but also paves the way for practical implementations that can drive inclusive and sustainable development.