The increase in demand for scalable and energy-efficient artificial neural networks has put the focus on novel hardware solutions. Integrated photonics offers a compact, parallel, and ultra-fast information processing platform, specially suited for extreme learning machine (ELM) architectures. Here we experimentally demonstrate a chip-scale photonic ELM based on wave chaos interference in a stadium microcavity. By encoding the input information in the wavelength of an external single-frequency tunable laser source, we leverage the high sensitivity to wavelength of injection in such photonic resonators. We fabricate the microcavity with direct laser writing of SU-8 polymer on glass. A scattering wall surrounding the stadium operates as a readout layer, collecting the light associated with the cavity's leaky modes. We report uncorrelated and aperiodic behavior in the speckles of the scattering barrier from a high-resolution scan of the input wavelength. Finally, we characterize the system's performance at classification in three qualitatively different benchmark tasks. As we can control the number of output nodes of our ELM by measuring different parts of the scattering barrier, we demonstrate the capability to optimize our photonic ELM's readout size to the performance required for each task.

Research on the prediction method of electro-thermal coupling thermal process of soft pack lithium-ion battery based on principal component analysis and extreme learning machine

SOH prediction for Lithium batteries using WPT and crested porcupine deep extreme learning machine under different temperatures

Estimation of uniaxial compressive strength of rock using optimized support vector and kernel-based extreme learning machine models

Rapid detection of deltamethrin content in drinking water based on the fusion strategy of Raman spectroscopy and UV-vis spectroscopy based on stoichiometric method.

Shannon entropy helps optimize the performance of a frequency-multiplexed extreme learning machine

A novel manifold discriminant extreme learning machine combined with an E-nose for chili pepper identification via aroma analysis

Research on optimization of library book recommendation system based on the collaborative fusion of transformer architecture and adaptive extreme learning machine

PLSELM: A lightweight modeling approach for low-data calibration in near-infrared spectroscopy.

Introduction Accurately determining the bottom boundary of anti-seepage curtains is critical for ensuring the integrity and performance of this key engineered composite structure in karst reservoirs. This study leverages artificial intelligence (AI) to address this materials design challenge. Methods We developed hybrid models by integrating a Genetic Algorithm (GA) with Backpropagation (BP), Support Vector Machine (SVM), and Extreme Learning Machine (ELM) algorithms. These models were trained and validated using a comprehensive dataset from the Dehou Reservoir, incorporating critical material and hydrogeological properties of the karst rock mass. A comparative analysis with Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM) was also conducted. Results The results demonstrated that GA optimization significantly enhanced predictive performance. The GA-BP model achieved superior accuracy (R 2 = 0.98, MSE = 7.58). Furthermore, from an engineering safety perspective, the GA-SVM model provided the most reliable recommendations, frequently yielding conservative depth estimates. The comparative analysis validated the competitive advantage of the proposed hybrid models over other benchmark algorithms. Discussion This research underscores the potential of AI-driven approaches for the performance prediction and rational design of engineered geomaterial systems. The findings offer a powerful tool for infrastructure projects in complex geological settings, balancing predictive accuracy with critical engineering safety considerations.

Introduction As the core equipment in industrial production, rotating machinery bearings play a critical role. However, traditional feature extraction algorithms for vibration signals are susceptible to noise interference and inaccurate in extracting complex features. Meanwhile, traditional fault classification algorithms face challenges such as high dependence on feature quality and insufficient generalization ability. Methods For vibration signal feature extraction, an improved multi-scale divergence entropy method is proposed. It integrates multi-scale sample entropy and divergence entropy to enhance the discrimination of signal features. For fault classification, a regularized extreme learning machine (ELM) model is developed, where regularization constraints are introduced to avoid pathological matrices. Results When using the refined composite multi-scale divergence entropy for feature extraction, setting the scale to 20 minimized the entropy value and achieved the highest classification accuracy of 98.79%. For the regularized ELM model, adopting the Softplus function as the activation function and setting the neuron number to 17 led to the lowest loss rate and the highest average classification accuracy of 93.98% ± 0.94%. Additionally, the model exhibited a relatively short running time of only 400 ms. Discussion The results indicate that the improved multi-scale divergence entropy effectively enhances the robustness and accuracy of feature extraction under noise interference. The regularized ELM model improves both classification accuracy and computational efficiency compared to traditional algorithms. This proposed method not only advances the classification accuracy of rotating machinery faults but also provides new technical support for machine fault prevention work, demonstrating potential for practical industrial applications in fault diagnosis systems.

Surface soil moisture (SSM) is a critical indicator of crop growth conditions, and its accurate retrieval is essential for agricultural monitoring. Integrating multispectral and microwave remote sensing data can enhance SSM estimation, but discrepancies among platforms often reduce accuracy at local scales. In this study, we fused Sentinel-2 and UAV multispectral images through resampling to generate fusion data, which were then combined with miniature synthetic aperture radar (MiniSAR) data. A modified water cloud model (WCM) was applied to mitigate vegetation effects on radar backscattering coefficients. Three machine learning algorithms—random forest (RF), extreme gradient boosting (XGBoost), and extreme learning machine (ELM)—were employed to retrieve SSM. Field measurements at two depths (0–10 cm and 0–20 cm) over winter wheat fields in Xunxian, Hebi City, Henan Province, China, were used for validation. Results showed the following: (1) Fused multispectral data improved retrieval accuracy compared with single-satellite data, with the best configuration (fused data + VV + RF) achieving an R2 of 0.85 and an RMSE of 1.51% at 0–10 cm. (2) At 0–20 cm, the fused data combined with VV polarization and XGBoost achieved the best performance (R2 = 0.67, RMSE = 2.61%). (3) ELM exhibited the largest accuracy improvement after incorporating fused data, with R2 increases up to 0.40 and RMSE reductions up to 18.24%. These results demonstrate the strong potential of multi-platform multispectral fusion combined with MiniSAR data for improving field-scale SSM retrieval in winter wheat regions.

ABSTRACT Uniaxial compressive strength (UCS) serves as a fundamental mechanical property for evaluating the failure resistance of rock. Traditional destructive tests, while accurate, are time-consuming, costly, and less suitable for soft or fractured rocks. To address these issues, this study proposes a novel non-destructive approach for lightweight and accurate UCS prediction by integrating characteristic impedance with deep learning. Five hybrid Deep Extreme Learning Machine (DELM) models were developed by adjusting hyperparameters via meta-heuristic algorithms, and their performance was evaluated on two databases containing characteristic impedance and P-wave velocity, respectively. Model interpretability was enhanced through Shapley Additive Explanations (SHAP) analysis. The results demonstrate that all optimised DELM models significantly outperformed than the unoptimised model. When using characteristic impedance database, the CTCM-DELM model could achieve the highest accuracy (R2 = 0.967 for training, R2 = 0.968 for testing). Note that models utilising characteristic impedance consistently exceeded those using P-wave velocity, indicating characteristic impedance as a more effective input parameter. SHAP analysis further identified characteristic impedance as the most influential variable in UCS prediction, followed by point load index and Schmidt hammer rebound value. This study confirmed that leveraging characteristic impedance and explainable data-driven model in UCS prediction could give high-fidelity UCS prediction value.

Precise short-term forecasting of photovoltaic (PV) power is essential for grid stability and the integration of renewables. We propose two hybrid architectures—TS-BERT+ELM and PatchTST+ELM—separate temporal representation learning from regression by integrating transformer-based encoders with a ridge-regularized Extreme Learning Machine (ELM) for rapid, low-latency prediction. An evaluation of one-day-ahead predictions from 14-day input windows is conducted using daily PV datasets from five EU nations (Germany, France, Switzerland, Denmark, and the United Kingdom) provided from OPSD and enhanced with NASA POWER meteorological variables (global horizontal irradiance, cloud cover, and temperature) (f : R 14×d → R). We present MAE, MSE, R2 , and threshold accuracies (Accuracy@10%, Accuracy@50%), cexecute ablation, convergence, and sensitivity studies, and conduct paired t-tests and Wilcoxon signed-rank tests for statistical validation. Results indicate that TS-BERT+ELM regularly surpasses baselines on noisy and irregular datasets (France, Germany), whereas PatchTST+ELM demonstrates strong performance with high-quality, structured data (Denmark, UK); Switzerland occupies a position bridging the two categories. Integrating external weather-related features further enhances predictive accuracy and decreases variance, with statistically significant gains (p < 0.05) in four countries and an inconclusive UK case due to high variance. This modular design facilitates rapid convergence, maintains robustness against missing inputs, and enhances operational efficiency, and is compatible with federated and transfer learning for privacy-preserving, cross-site deployment. These findings support scalable, multimodal, and privacy-aware PV forecasting in real-world energy systems.

Real-time monitoring of rock stability and effective control of pressure concentration areas are crucial for ensuring the safety of personnel and equipment during mineral resource extraction. Microseismic and blasting signals, as early indicators of rock rupture, can effectively predict potential disasters. This study proposes a binary Harris hawks optimization algorithm with kernel search (bKSHHO) for the recognition of microseismic and blasting signal data. By integrating bKSHHO with a kernel extreme learning machine (KELM), we construct a prediction model, termed bKSHHO-KELM, which predicts microseismic and blasting signals, thereby enabling early warning of rock hazards. Experimental studies validate the optimization capability of the proposed KSHHO algorithm by comparing it with ten peer algorithms using the IEEE CEC 2022 benchmark functions. The bKSHHO-KELM method was then applied to predict microseismic and blasting signals, achieving an accuracy of 95.625%, a recall of 93.964%, a precision of 92.632%, and an F1 score of 0.931. This provides an efficient and accurate early warning solution for microseismic hazards in mine safety management.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-24211-4.

Abstract Accurate forecasting of foreign exchange (Forex) rates is critical for financial decision-making, given the nonlinear and stochastic nature of market data. This study proposes a hybrid forecasting model, ELM-GA-ERWCA, which integrates extreme learning machine (ELM) with genetic algorithm (GA) and evaporation rate water cycle algorithm (ERWCA). Three currency pairs, USD:INR, SAR:INR, and SGD:INR, were analyzed using 4000 daily samples (2005–2020). Datasets were reconstructed with technical indicators and evaluated in both segregated and un-segregated forms. Performance was assessed using RMSE, MAPE, and R 2 across short- and long-term prediction horizons. Results show that the proposed model consistently outperforms baseline models (ELM-GA, ELM-WCA, ELM-ERWCA), achieving RMSE reductions of up to 12%, MAPE improvements of 8–10%, and R 2 values above 0.99. Convergence analysis confirmed faster and more stable optimization, while Friedman statistical validation established the robustness of the approach. The findings demonstrate that ELM-GA-ERWCA provides a statistically reliable framework for Forex prediction, with potential for future integration into trading simulations and risk-aware financial applications. The proposed ELM-GA-ERWCA model demonstrates statistically robust forecasting accuracy across multiple currency datasets. Its lower error margins and consistent convergence behavior indicate potential for practical application in financial decision support systems. However, its economic implications must be further validated through trading simulations and backtesting frameworks before being considered a risk-minimizing tool for investors. Figure A gives complete idea about the summary of the work. Graphical Abstract

Autism is a complex psychiatric condition that needs to be diagnosed early using more objective techniques. Hence, many researchers have turned to diagnosing autism by analyzing EEG signals. However, a comprehensive framework for this has not yet been introduced, and there is room for improvement. In this study, to increase the precision of autism diagnosis from EEG signals, a new framework is introduced that includes the steps of data preprocessing, extraction of nonlinear features from EEG time series, optimization of extracted features using an innovative technique based on GA and DBSCAN algorithms, feature reduction using the LASSO technique, and classification using a fuzzy ELM classifier. This study presents a feature optimization technique that leverages a genetic algorithm informed by clustering principles. Rather than relying on random selection for forming the new generation, the approach incorporates clustering during the fitness evaluation phase to identify and exclude outliers from advancing to the next generation. The recommended scheme was examined on two EEG databases. Using only 14-channel EEG data, it was able to achieve 96.81% accuracy, 95.16% sensitivity, 97.73% specificity, and a 96.42% F1-score for autism detection using database A, as well as 97.64% accuracy, 96.55% sensitivity, 98.49% specificity, and a 97.51% F1-score using database B. This framework outperformed existing methods on two EEG databases. The practical use of low-channel systems suggests potential for real-world clinical deployment, enabling scalable and cost-effective screening. This study underscores the potential of using nonlinear dynamics of EEG signals alongside fuzzy ELM for diagnosing autism in children.

Abstract The application of Machine Learning (ML) models coupled with metaheuristic optimization algorithms represents a potentially powerful development in the field of predictive modeling, as it relates to sustainable energy materials. In this study, the electrochemical performance of biomass material from bamboo for energy storage applications is explored, focusing on the prediction of power density. The central objective is to enhance model accuracy with a novel hybrid model of Kernel Extreme Learning Machine (KELM) and Slime Mould Algorithm (SMA), Sunflower Optimization (SFO), and Social Ski Driver (SSD). The optimal predictive performance was achieved with KELM-SFO with a test Root Mean Square Error (RMSE) of 10,421.05, Mean Absolute Error (MAE) of 4,654.07, and R-squared (R 2 ) of 96.2 %. Early and fast plateauing of the SFO algorithm’s convergence curve indicated stable, early-stage optimization. In addition to filling a significant knowledge gap in ML-incorporated materials modeling, this work opens the door for future research on deep learning techniques, adaptive hybrid optimization algorithms, and real-time experimental validations to improve the electrochemical prediction efficiency in energy storage systems inspired by biomass.

The reliable and safe operation of traction elevators depends on traction capacity, which is degraded by traction sheave groove wear. The resulting slippage reduces transmission efficiency and may cause a catastrophic failure due to the sudden loss of friction. After analyzing slippage mechanisms, we propose a prediction model that combines the Improved Pelican Optimization Algorithm (IPOA) with an Extreme Learning Machine (ELM). A mechanism analysis identifies key inputs—the wear amount, payload, and wire rope tension—providing a basis for model construction. The approach uses Halton sequence initialization, adaptive nonlinear weighting, and Gaussian perturbation, which improve the handling of nonlinearities. IPOA is then employed to optimize the ELM parameters, yielding the IPOA-ELM model. Experiments across multiple wear conditions show that IPOA-ELM predicts slippage more accurately than a traditional ELM. The study clarifies how traction sheave groove wear induces rope slippage and demonstrates the effectiveness of the proposed model under varying wear and load conditions, offering a practical reference for failure mechanism analysis and preventive strategies in elevator traction systems.

ABSTRACT Accurate prediction of river water quality parameters is essential for environmental protection and sustainable agricultural resource management. This study presents a novel framework for estimating potential salinity in river water in arid and semi‐arid regions by integrating a kernel extreme learning machine (KELM) with a boosted salp swarm algorithm based on differential evolution (KELM‐BSSADE). A dataset of 336 samples, including bicarbonate, calcium, pH, total dissolved solids and sodium adsorption ratio, was collected from the Idenak station in Iran and was used for the modelling. Results demonstrated that KELM‐BSSADE outperformed models such as deep random vector functional link (dRVFL), general regression neural network (GRNN), multivariate adaptive regression spline (MARS), online sequential extreme learning machine (OSELM) and extreme gradient boosting decision tree (XGBoost) when compared with observed river salinity data. Also, the KELM‐BSSADE model effectively identified optimal inputs through the Boruta‐XGBoost (B‐XGB) feature selection method. Four metaheuristic‐based KELM models were developed, utilizing grey wolf optimizer, whale optimization, slime mould algorithm and equilibrium optimizer, further illustrating the capability of KELM‐BSSADE in estimating potential salinity in river water. By accurately estimating potential salinity, KELM‐BSSADE can assist in optimizing irrigation practices, ensuring that agricultural demands are met while minimizing the risk of salinity‐related crop damage.