Article
Mathematics
Hau T. Mai, Jaewook Lee, Joowon Kang, H. Nguyen-Xuan, Jaehong Lee
Summary: This paper presents an improved surrogate blind Kriging (IBK) and a combined infill strategy to solve constrained expensive black-box optimization problems. By enhancing the prediction accuracy of metamodels and updating them with an infill strategy, IBK can efficiently find the optimal solution and has been successfully applied to structural design optimization.
Article
Engineering, Civil
Aaron Cardenas-Martinez, Victor Rodriguez-Galiano, Juan Antonio Luque-Espinar, Maria Paula Mendes
Summary: This study evaluates the performance of machine learning algorithms combined with feature selection techniques in predicting nitrate pollution, using different environmental features. By analyzing the dynamics of agricultural activity and changes in nitrate pollution, key features related to nitrate pollution can be identified.
JOURNAL OF HYDROLOGY
(2021)
Article
Agronomy
Gunnar Lischeid, Heidi Webber, Michael Sommer, Claas Nendel, Frank Ewert
Summary: This study utilized machine learning methods to investigate the impact of climatic and soil hydrological factors on the yield of four crops, highlighting the uniqueness of key predictors. Random Forest and Support Vector Machine models achieved between 50% and 70% capture of spatial and temporal variance, with different sets of predictors performing similarly. In light of climate change, excess precipitation and heat effects are seen as important factors in crop breeding and modeling.
AGRICULTURAL AND FOREST METEOROLOGY
(2022)
Article
Computer Science, Interdisciplinary Applications
Nicolau Andres-Thio, Mario Andres Munoz, Kate Smith-Miles
Summary: Multifidelity expensive black-box methods have gained attention for their applicability in industrial design problems. The challenge lies in the limited knowledge of the relationship between decisions and objective values. Surrogate models have been developed to address this issue. This study expands the existing test instances and explores the reliability of low-fidelity information sources.
INFORMS JOURNAL ON COMPUTING
(2022)
Article
Multidisciplinary Sciences
Cheng-Han Chua, Meihui Guo, Shih-Feng Huang
Summary: This paper proposes a KC Score to measure feature importance in clustering analysis of high-dimensional data. The KC Score-based feature selection strategy effectively selects important features for clustering, achieving the same or better clustering performance compared to using all features in several datasets.
SCIENTIFIC REPORTS
(2022)
Article
Automation & Control Systems
Jun Ma, Chunyang Yin, Xiaoke Li, Xinyu Han, Wuyi Ming, Shiyou Chen, Yang Cao, Kun Liu
Summary: In order to balance the cost and accuracy in optimizing the EDM process parameters, a variable-fidelity surrogate model (VFM) was introduced to replace the implicit relationship between the process parameters and performance functions. By combining low-fidelity and high-fidelity samples, an accurate expression of the relationship between process parameters and performance functions was established. The results of the optimization experiments showed that the parameter combination obtained by VFM outperformed both the low-fidelity model and the high-fidelity model, achieving a higher MRR while meeting the constraint on surface roughness.
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
(2022)
Article
Computer Science, Information Systems
Xiangyuan Gu, Jianguo Chen, Guoqiang Wu, Kun Wang, Jiaxing Wang
Summary: Due to the limitation of using only one metric or comparing two metrics separately to measure redundant features, some existing feature subset selection algorithms fail to achieve the desired performance. To address this issue, a feature subset selection algorithm called symmetric uncertainty and interaction factor (SUIF) is proposed. SUIF evaluates relevant features using symmetric uncertainty and removes irrelevant features. It then uses graph theory to process relevant features, removes edges with lower weights, and clusters features using the Louvain community detection algorithm. Finally, SUIF evaluates features in each cluster using symmetric uncertainty, interaction factor, and equal interval division and ranking, and eliminates redundant features. Experimental results show that SUIF outperforms other algorithms in terms of feature selection performance.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Interdisciplinary Applications
Qi Guo, Jiutao Hang, Suian Wang, Wenzhi Hui, Zonghong Xie
Summary: This paper presents an efficient design optimization method assisted by multi-fidelity surrogate models for buckling design of variable stiffness composites. By using hierarchical Kriging and global optimization method, the effectiveness and robustness of the method are demonstrated through two case studies.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2021)
Article
Computer Science, Artificial Intelligence
Yingyue Chen, Yumin Chen
Summary: By introducing neighborhood rough set model and variable precision neighborhood rough set model, this paper addresses the issues of handling real-value data and weak fault tolerance in traditional rough sets. The proposed method enhances the fault-tolerant ability of classification systems and designs an algorithm to select feature subsets. Experimental results demonstrate the effectiveness and compactness of the feature subset selection.
INTERNATIONAL JOURNAL OF COMPUTATIONAL INTELLIGENCE SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Zhi Jiang, Yong Zhang, Jun Wang
Summary: The paper proposes a new ensemble feature selection algorithm, MDEFS, which can handle large-scale data, reduce computational costs, and improve the accuracy of feature selection results.
APPLIED SOFT COMPUTING
(2021)
Review
Computer Science, Artificial Intelligence
Carlos Villa-Blanco, Concha Bielza, Pedro Larranaga
Summary: Real-world problems often have high feature dimensionality, making it difficult to model and analyze the data. Feature subset selection (FSS) techniques can be used to reduce irrelevant or redundant information, improving the speed and performance of building models. This review focuses on incremental FSS algorithms that can efficiently handle large volumes of data received sequentially. Different strategies, such as updating feature weights incrementally, applying information theory, or using rough set-based FSS, are discussed, along with various supervised and unsupervised learning tasks where FSS is applicable.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Engineering, Mechanical
Umberto Alibrandi, Lars V. Andersen, Enrico Zio
Summary: This paper applies information theory to probabilistic sensitivity analysis and surrogate modelling with active learning. It introduces a new measure of dependence between random variables using the informational coefficient of correlation. The paper also presents effective informational sensitivity indices based on mutual information and proposes two novel learning functions for adaptive sampling. Numerical examples demonstrate the features and potential applications of the proposed approach.
PROBABILISTIC ENGINEERING MECHANICS
(2022)
Article
Automation & Control Systems
Daniel R. Kowal
Summary: Subset selection is a valuable tool for interpretability, scientific discovery, and data compression. We propose a Bayesian approach to address the challenges in classical subset selection, and introduce a strategy that focuses on finding near-optimal subsets rather than a single best subset. We apply Bayesian decision analysis to derive the optimal linear coefficients for any subset of variables, and our approach outperforms competing methods in prediction, interval estimation, and variable selection. By analyzing a large education dataset, we gain unique insights into the factors that predict educational outcomes and identify over 200 distinct subsets of variables that offer near-optimal predictive accuracy.
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Materials Science, Multidisciplinary
Debiao Meng, Yan Li, Chao He, Jinbao Guo, Zhiyuan Lv, Peng Wu
Summary: An enhanced CO method based on adaptive surrogate models is proposed to improve the accuracy of the non-linear region of the response surface. The method modifies traditional surrogate models and replaces original objectives and constraints with adaptive surrogate models to enhance the optimization process. This approach demonstrates the effectiveness of CO in engineering structure design problems.
MATERIALS & DESIGN
(2021)
Article
Engineering, Industrial
Jiayi Ding, Jianfang Zhou, Wei Cai
Summary: This paper proposes an efficient variable selection-based Kriging model method to approximate the finite element analysis model in reliability analysis of slopes. The variable selection technique successfully solves the curse of dimensionality problem within Kriging model induced by numerous random variables. The implementation procedure of this method for the reliability analysis of slopes is introduced in detail, and the validity is demonstrated through examples.
RELIABILITY ENGINEERING & SYSTEM SAFETY
(2023)
Article
Computer Science, Interdisciplinary Applications
Nasrulloh Loka, Ivo Couckuyt, Federico Garbuglia, Domenico Spina, Inneke Van Nieuwenhuyse, Tom Dhaene
Summary: Multi-objective optimization of complex engineering systems is a challenging problem. Bayesian optimization is a popular technique to tackle this problem. We develop an approach that can handle a mix of expensive and cheap objective functions, offering lower complexity and superior performance in cases where the cheap objective function is difficult to approximate.
ENGINEERING WITH COMPUTERS
(2023)
Article
Construction & Building Technology
Janez Perko, Eric Laloy, Rafael Zarzuela, Ivo Couckuyt, Ramiro Garcia Navarro, Maria J. Mosquera
Summary: The effectiveness of sol-gel based treatments for concrete protection depends on their ability to penetrate the pores of the material. Optimizing the sol formulation to achieve maximum penetration depth is complex due to the varying influence of sol's physical properties with different concrete pore size distributions. This manuscript presents a combined computational and experimental approach to design impregnation products with optimized penetration depth on concrete with different pore structures. The effectiveness of the approach is demonstrated through three cases, showing significant improvement in penetration compared to traditional methods.
CEMENT & CONCRETE COMPOSITES
(2023)
Article
Geochemistry & Geophysics
Geethika Bhavanasi, Lorin Werthen-Brabants, Tom Dhaene, Ivo Couckuyt
Summary: This paper builds and compares open-set recognition (OSR) systems for patient activity recognition (PAR) using compact radar sensors in a hospital setting. A deep discriminative representation network (DDRN) is trained using large margin cosine loss (LMCL) and triplet loss (TL), and a probability of an inclusion model based on the Weibull distribution is used to separate knowns from unknowns. The proposed approach significantly outperforms state-of-the-art open-set approaches for human activity recognition (HAR) with radar.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Operations Research & Management Science
Jixiang Qing, Ivo Couckuyt, Tom Dhaene
Summary: Bayesian optimization is a popular tool for optimizing objective functions with limited function evaluations. This paper introduces a novel Bayesian optimization framework for multi-objective optimization considering input uncertainty. The framework utilizes a robust Gaussian Process model and a two-stage optimization process to find robust solutions.
JOURNAL OF GLOBAL OPTIMIZATION
(2023)
Article
Engineering, Electrical & Electronic
Federico Garbuglia, Domenico Spina, Dirk Deschrijver, Ivo Couckuyt, Tom Dhaene
Summary: In microwave design, Bayesian optimization techniques are used to optimize the frequency response of components and devices. This article proposes using a deep Gaussian process to directly model all relevant S coefficients over the frequency and design parameter ranges of interest, leading to increased accuracy in identifying the optimal frequency response.
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES
(2023)
Article
Chemistry, Analytical
Laurens D'hooge, Miel Verkerken, Tim Wauters, Filip De Turck, Bruno Volckaert
Summary: Recently proposed intrusion detection methods are being tested and validated on a sparse collection of academic datasets. However, the meaningfulness of the improvements demonstrated on these datasets is rarely investigated. This study aims to show that strong classification performance on current intrusion detection datasets does not necessarily translate to strong general performance.
Article
Water Resources
Michael Weyns, Ganjour Mazaev, Guido Vaes, Filip Vancoillie, Filip De Turck, Sofie Van Hoecke, Femke Ongenae
Summary: Water loss due to persistent leakages in water distribution networks is a significant problem worldwide, especially with increasing global water scarcities. This paper presents a data-driven approach using a connected Geographical Information System and an autoencoder for anomaly detection on time-variable sensor data to localize leaks. The approach exploits leakless data during model training, eliminating the need for a large variety of leak scenarios. Based on field evaluations with 19 realistic leak experiments, the approach achieved average search costs as low as 2.2 kilometers for a total network length of 215 kilometers.
URBAN WATER JOURNAL
(2023)
Article
Computer Science, Information Systems
Yuan-Cheng Lai, Didik Sudyana, Ying-Dar Lin, Miel Verkerken, Laurens D'hooge, Tim Wauters, Bruno Volckaert, Filip De Turck
Summary: This study adopts an ML-based IDS with multi-tier architecture and task assignments to improve the capability of detecting network intrusions. Through evaluation using queueing theory and simulated annealing, it was found that task assignments on the edge node achieved the best performance. Additionally, a two-tier architecture with edge and cloud components significantly reduced the delay for IDS as a Service.
IEEE TRANSACTIONS ON NETWORK AND SERVICE MANAGEMENT
(2023)
Article
Health Care Sciences & Services
Stephanie Carlier, Vince Naessens, Femke De Backere, Filip De Turck
Summary: The use of serious games in health care is increasing, but current games lack personalized interventions and require extensive development and multidisciplinary involvement. This study proposes a software engineering framework to streamline the design process of personalized serious games in health care, focusing on knowledge transfer and algorithm reuse. The framework assigns responsibilities to stakeholders and simplifies the design process by answering three key questions about personalization.
JMIR SERIOUS GAMES
(2023)
Article
Computer Science, Information Systems
Mays AL-Naday, Nikolaos Thomos, Jiejun Hu, Bruno Volckaert, Filip de Turck, Martin J. Reed
Summary: Digital transformation relies on service-based operations in fog networks, which extend cloud resources closer to end-users. This dispersion of resources enables diversity and coexistence of multiple providers, but also leads to variation in operational cost and limited information sharing. To address this, a novel service-based fog management and network orchestrator (sbMANO) is proposed, which utilizes service metadata for multi-provider resource management and optimization.
IEEE TRANSACTIONS ON SERVICES COMPUTING
(2023)
Article
Computer Science, Information Systems
Hemanth Kumar Ravuri, Maria Torres Vega, Jeroen Van Der Hooft, Tim Wauters, Filip De Turck
Summary: The increasing popularity of HMD and depth cameras has led to the demand for interactive immersive media content over the internet. Traditional video streaming using DASH over HTTP suffers from latency and playout interruptions. HTTP/3 replaces TCP with QUIC at the transport layer to solve these issues and allows for unreliable data delivery, reducing latency. This work proposes integrating DASH with partial reliability of QUIC to reduce interruptions and enhance the quality of immersive content delivery on lossy networks.
Article
Engineering, Electrical & Electronic
Federico Garbuglia, Torsten Reuschel, Christian Schuster, Dirk Deschrijver, Tom Dhaene, Domenico Spina
Summary: This work presents a machine learning technique using a new Gaussian processes model to accurately model wide-band scattering parameters of interconnects. By employing delay estimation and a physics-informed kernel, the new model accurately predicts S-parameter values and outperforms standard models in terms of accuracy.
IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY
(2023)