Article
Meteorology & Atmospheric Sciences
C. Pelaez-Rodriguez, J. Perez-Aracil, C. Casanova-Mateo, S. Salcedo-Sanz
Summary: Low visibility events pose a serious problem for road transport, leading to accidents and economic losses. Accurately predicting these events can help prevent such issues. Machine and deep learning techniques have been employed to predict fog using in situ meteorological data and persistence variables. The study evaluates these techniques for different prediction time-horizons and investigates the inclusion of data from ERA5 Reanalysis. An iterative forward selection algorithm based on evolutionary algorithms is proposed to determine optimal variables and nodes, resulting in improvements in prediction accuracy.
ATMOSPHERIC RESEARCH
(2023)
Article
Meteorology & Atmospheric Sciences
C. Castillo-Boton, D. Casillas-Perez, C. Casanova-Mateo, S. Ghimire, E. Cerro-Prada, P. A. Gutierrez, R. C. Deo, S. Salcedo-Sanz
Summary: This paper provides a comprehensive analysis of low-visibility event prediction problems formulated as both regression and classification tasks. It discusses the performance of various ML approaches and evaluates their performance under a common comparison framework. The results can guide the selection of the most efficient formulation and best performing ML approaches for low-visibility event prediction.
ATMOSPHERIC RESEARCH
(2022)
Article
Computer Science, Artificial Intelligence
C. Pelaez-Rodriguez, J. Perez-Aracil, A. de Lopez-Diz, C. Casanova-Mateo, D. Fister, S. Jimenez-Fernandez, S. Salcedo-Sanz
Summary: In this paper, different Deep Learning-based ensemble algorithms are proposed and discussed for predicting low-visibility events due to fog. Seven different Deep Learning architectures are considered, generating multiple individual learners. The models' hyperparameters are randomly selected within a pre-defined range, and each model is trained with slightly different data. Three information fusion techniques are employed to build the ensemble models, and the influence of filtering process and elitism level is assessed. The proposed methodology shows good performance compared to other Machine Learning, DL algorithms, and meteorological-based methods in low-visibility events prediction.
Article
Automation & Control Systems
Jiahang Liu, Lei Zuo, Xin Xu, Xinglong Zhang, Junkai Ren, Qiang Fang, Xinwang Liu
Summary: This paper introduces a novel batch-mode RL approach with randomly projected features for VFA, named ELM-API, which outperforms previous API algorithms with advantages such as quick feature generation and adaptability. Through comprehensive simulation studies on benchmark learning control problems and a challenging high-dimensional lane-changing decision problem, the effectiveness of ELM-API in real-world applications is demonstrated.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2021)
Article
Physics, Multidisciplinary
L. Innocenti, S. Lorenzo, I. Palmisano, A. Ferraro, M. Paternostro, G. M. Palma
Summary: Quantum machine learning applies machine learning concepts and techniques to quantum devices. This paper presents a framework to model quantum extreme learning machines (QELMs), demonstrating that they can be described concisely through single effective measurements and characterizing the information retrievable using such protocols. Understanding the capabilities and limitations of QELMs will enable their full deployment in system identification, device performance optimization, and state or process reconstruction.
COMMUNICATIONS PHYSICS
(2023)
Article
Computer Science, Artificial Intelligence
A. L. Afzal, Nikhitha K. Nair, S. Asharaf
Summary: The emergence of extreme learning machine as a rapid learning algorithm has marked its prominence in solitary hidden layer feed-forward networks. This paper introduces a deep kernel learning approach in a conventional shallow architecture and explores the possibility of building a new deep kernel machine with extreme learning machine and multilayer arc-cosine kernels.
PATTERN ANALYSIS AND APPLICATIONS
(2021)
Article
Computer Science, Information Systems
D. Anitha, R. A. Karthika
Summary: In this article, a novel routing scheme is proposed based on Q-learning framework and Deep Extreme Learning Machines aided by Adaptive Firefly Routing algorithm to address energy efficiency and network instability issues in underwater communication. The proposed algorithm outperformed existing algorithms in terms of complexity, energy consumption, packet delivery ratio, and end to end delay through extensive experimentation.
COMPUTER COMMUNICATIONS
(2021)
Article
Meteorology & Atmospheric Sciences
Erica F. De Biasio, Konstantine P. Georgakakos
Summary: Using a high-resolution WRF mesoscale model, this study explores the upstream atmospheric precursors that contribute to the enhancement of precipitation over the mountain regions of Southern California. The researchers find statistically significant physics-based signals between hypothetical mesoscale forcings and the modeled precipitation response. The presence of convective available potential energy (CAPE) indicates the role of atmospheric instability in facilitating high short-duration precipitation intensities.
JOURNAL OF HYDROMETEOROLOGY
(2023)
Article
Water Resources
Trevor Page, Keith J. J. Beven, Barry Hankin, Nick A. A. Chappell
Summary: The study compared different interpolation methods for rainfall interpolation during extreme rainfall events, showing that using the CK method with an effective elevation index as a secondary variable performed the best, with an overall improvement of around 40%.
HYDROLOGICAL PROCESSES
(2022)
Article
Multidisciplinary Sciences
Murad Al-Rajab, Samia Loucif, Yazan Al Risheh
Summary: The world's population is expected to grow 32% in the coming years, with the number of Muslims projected to increase by 70% from 1.8 billion in 2015 to about 3 billion in 2060. The Hijri calendar, based on lunar months, is used by Muslims to determine important dates and events. However, there is no consensus within the Muslim community on the start of Ramadan due to imprecise observations of the new crescent Moon. This paper proposes the use of machine learning algorithms, particularly Random Forest and Support Vector Machine, to accurately predict the visibility of the new Moon and determine the start of Ramadan.
SCIENTIFIC REPORTS
(2023)
Article
Automation & Control Systems
XuDong Shi, Qi Kang, Jing An, MengChu Zhou
Summary: This article proposes a novel L1 norm-based extreme learning machine (ELM) by integrating bound optimization theory with variational Bayesian inference. The proposed method efficiently solves the overfitting problem and demonstrates competitive performance in an industrial case study.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2022)
Article
Chemistry, Multidisciplinary
Francesco Mercaldo, Luca Brunese, Fabio Martinelli, Antonella Santone, Mario Cesarelli
Summary: Currently, deep learning networks, particularly convolutional neural network models, are commonly used for biomedical image classification. However, deep learning is expensive to train due to complex data models. A recent alternative called extreme learning machine has been shown to achieve acceptable predictive performance in classification tasks at a much lower training cost compared to deep learning networks. In this study, we explore the potential of using extreme learning machines for biomedical classification tasks and demonstrate their effectiveness in binary and multiclass classification of biomedical images acquired with the dermatoscope and blood cell microscope.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Information Systems
Denis Reilly, Mark Taylor, Paul Fergus, Carl Chalmers, Steven Thompson
Summary: This paper addresses the challenges of handling purely categorical datasets and provides a set of heuristics for dealing with such data. A case study using a dataset of domestic fire injuries is presented to demonstrate the effectiveness of the heuristics and analyze the performance of different encoding techniques. The paper makes novel contributions in solving purely categorical dataset problems and classifying injury types.
Article
Computer Science, Artificial Intelligence
Megan Boucher-Routhier, Bill Ling Feng Zhang, Jean-Philippe Thivierge
Summary: The article presents a rapid one-shot learning rule for training recurrent neural networks to reproduce natural images, sequential patterns, and high-resolution movie scenes. This approach offers a new avenue for one-shot learning in biologically realistic recurrent networks.
Article
Computer Science, Artificial Intelligence
Fabricio Olivetti de Franca, Maira Zabuscha de Lima
Summary: Symbolic Regression is a technique that searches for a mathematical expression fitting an input data set by minimizing approximation error, exploring a search space filled with redundancy and ruggedness. The new representation called Interaction Transformation can model function forms as a composition of interactions and a transformation function, showing promising results with lower computational cost compared to the current state-of-the-art.
Article
Green & Sustainable Science & Technology
Ravinesh C. Deo, A. A. Masrur Ahmed, David Casillas-Perez, S. Ali Pourmousavi, Gary Segal, Yanshan Yu, Sancho Salcedo-Sanz
Summary: Prediction of Total Cloud Cover (TCDC) from numerical weather simulation models can aid renewable energy engineers in monitoring and forecasting solar photovoltaic power generation. A major challenge is the systematic bias in TCDC simulations induced by errors in the numerical model parameterization stages. Correction of GFS-derived cloud forecasts at multiple time steps can improve energy forecasts in electricity grids to bring better grid stability or certainty in the supply of solar energy.
Article
Computer Science, Artificial Intelligence
Sujan Ghimire, Thong Nguyen-Huy, Ramendra Prasad, Ravinesh C. Deo, David Casillas-Perez, Sancho Salcedo-Sanz, Binayak Bhandari
Summary: This study proposes a hybrid method that combines convolutional neural network (CNN) with multi-layer perceptron (MLP) to generate solar radiation forecasts. The proposed CMLP model shows excellent performance in predicting solar radiation at various study sites. It should be explored as a viable modelling tool for real-time energy management systems.
COGNITIVE COMPUTATION
(2023)
Article
Thermodynamics
C. G. Marcelino, G. M. C. Leite, E. F. Wanner, S. Jimenez-Fernandez, S. Salcedo-Sanz
Summary: This study proposes a modeling and optimization method for a hybrid microgrid system (HMGS) using a swarm-intelligent algorithm and a Net-Metering compensation policy. Real industrial and residential data from a Spanish region are used to analyze the impact of four different Net-Metering compensation levels on costs, percentage of renewable energy sources (RESs), and LOLP. The results show that the Net-Metering policy reduces surplus and increases RESs participation in the microgrid, and using a VRFB system with a 25% compensation policy can yield significant cost savings compared to using a LTO system without Net-Metering.
Article
Biochemistry & Molecular Biology
Lucas Cuadra, Sancho Salcedo-Sanz, Jose Carlos Nieto-Borge
Summary: This paper presents a network model for studying the transport behavior of electrons and holes in IB solar cells, and proposes a design constraint of reducing carrier effective mass and inter-dot distance to increase transport efficiency.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2023)
Article
Green & Sustainable Science & Technology
G. M. C. Leite, C. G. Marcelino, C. E. Pedreira, S. Jimenez-Fernandez, S. Salcedo-Sanz
Summary: In recent years, distributed clean energy resources such as solar irradiation, wind generation, and electric vehicles have been considered as alternatives to fossil fuels. Managing these resources is complex due to their uncertain behavior. This study addresses a risk-based energy resource management optimization problem and proposes an improved version of Canonical Differential Evolutionary Particle Swarm Optimization (C-DEEPSO) to solve it. The results show that the proposed C-DEEPSO can provide solutions that reduce costs and protect against extreme scenarios.
JOURNAL OF CLEANER PRODUCTION
(2023)
Article
Green & Sustainable Science & Technology
Lionel P. Joseph, Ravinesh C. Deo, Ramendra Prasad, Sancho Salcedo-Sanz, Nawin Raj, Jeffrey Soar
Summary: This research proposes a novel hybrid bidirectional LSTM model for near real-time wind speed forecasting. The model utilizes wind speed and selected climate indices to predict wind speed, and applies a 3-stage feature selection to extract significant input features. The proposed hybrid BiLSTM algorithm outperforms other tested algorithms in wind speed prediction.
Review
Geochemistry & Geophysics
D. Barriopedro, R. Garcia-Herrera, C. Ordonez, D. G. Miralles, S. Salcedo-Sanz
Summary: Heat waves have significant socioeconomic and environmental impacts, and their frequency, intensity, and duration are projected to increase with global warming. While some thermodynamic processes have been identified, there is still a lack of understanding regarding dynamical aspects, regional forcings, and feedbacks, as well as their future changes.
REVIEWS OF GEOPHYSICS
(2023)
Article
Computer Science, Artificial Intelligence
C. G. Marcelino, J. Perez-Aracil, E. F. Wanner, S. Jimenez-Fernandez, G. M. C. Leite, S. Salcedo-Sanz
Summary: In this paper, a new hybrid optimization algorithm CE+CRO-SL is proposed to solve the optimal power flow problem. It outperforms traditional methods in terms of efficiency and accuracy, and achieves millions of dollars in profit in the tested scenarios.
Review
Chemistry, Multidisciplinary
Mehrdad Razzaghian Ghadikolaee, Elena Cerro-Prada, Zhu Pan, Asghar Habibnejad Korayem
Summary: Three-dimensional printed concrete has revolutionized the construction industry and there have been many studies to improve its performance. This study presents the main design properties of 3D printed concrete and covers the fresh and hardened state characteristics of concrete with different nano- and micro-additives.
Article
Mathematics
Jorge Perez-Aracil, Carlos Camacho-Gomez, Eugenio Lorente-Ramos, Cosmin M. Marina, Laura M. Cornejo-Bueno, Sancho Salcedo-Sanz
Summary: This paper proposes new probabilistic and dynamic strategies for creating multi-method ensembles based on the CRO-SL algorithm. Two different probabilistic strategies are analyzed to improve the algorithm. The performances of the proposed ensembles are tested for different optimization problems, comparing the results with existing algorithms in the literature.
Article
Computer Science, Interdisciplinary Applications
Salvin S. Prasad, Ravinesh C. Deo, Sancho Salcedo-Sanz, Nathan J. Downs, David Casillas-Perez, Alfio V. Parisi
Summary: This research aims to design an artificial intelligence-inspired early warning tool for short-term forecasting of UV index (UVI) in Australian hotspots. The proposed model outperformed all benchmarked models and its predictions are influenced by factors such as ozone effect and cloud conditions. The UVI prediction system reaffirms its benefits for providing real-time UV alerts and reducing health complications.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2023)
Article
Chemistry, Multidisciplinary
L. Cornejo-Bueno, J. Perez-Aracil, C. Casanova-Mateo, J. Sanz-Justo, S. Salcedo-Sanz
Summary: This study proposes a methodology based on classification and regression techniques to predict the occurrence and quantity of desert locusts. Different machine learning algorithms, such as linear regression, Support Vector Machines, decision trees, random forests, and neural networks, were applied and evaluated in Western Africa, primarily Mauritania. The results show that the random forest algorithm performs exceptionally well in both classification and regression tasks, making it the most effective machine learning algorithm among those used.
APPLIED SCIENCES-BASEL
(2023)
Article
Acoustics
Alberto Palomo-Alonso, David Casillas-Perez, Silvia Jimenez-Fernandez, Jose A. Portilla-Figueras, Sancho Salcedo-Sanz
Summary: This article proposes a flexible architecture with different algorithms for effective story segmentation of broadcast news from subtitle files. The proposed system uses spatial and temporal distance, as well as sentence similarity, to classify different stories in news broadcasts. The computational algorithms focus on each sentence's features and are combined to build an overall classifier. The proposed approach is evaluated using Video Text Track (VTT) subtitle files, and the algorithms are designed to handle noisy content.
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING
(2023)
Article
Engineering, Electrical & Electronic
Sara Garcia-de-Villa, David Casillas-Perez, Ana Jimenez-Martin, Juan Jesus Garcia-Dominguez
Summary: Inertial motion analysis has gained increasing interest due to its advantages over classical optical systems. This review examines the various proposals for inertial motion analysis found in the literature, including the publishers, sensors used, applications, algorithms, study participants, and validation systems employed. The review also explores recent developments in machine learning techniques and approaches to reduce estimation error. Overall, this paper provides an overview of the research in this field and suggests future research directions.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)