Article
Chemistry, Physical
Claudio Pereti, Kevin Bernot, Thierry Guizouarn, Frantisek Laufek, Anna Vymazalova, Luca Bindi, Roberta Sessoli, Duccio Fanelli
Summary: We propose an approach based on DeepSet technology for supervised classification and regression of superconductive materials. The method takes the chemical constituents as input, avoiding artefacts from ordering in the list. Successful performance is achieved in classifying superconducting materials and quantifying their critical temperature. Using the trained neural network, we searched the International Mineralogical Association list and identified three superconducting candidates, confirming superconductivity in the synthetic analogue of michenerite and observing it for the first time in monchetundraite with critical temperatures in good agreement with theory predictions. This marks the first certified superconducting material identified through artificial intelligence methodologies.
NPJ COMPUTATIONAL MATERIALS
(2023)
Article
Computer Science, Artificial Intelligence
Haishan Ye, Luo Luo, Zhihua Zhang
Summary: The novel algorithm APSSN introduces acceleration techniques to improve computational efficiency of the Newton-type proximal method while maintaining a fast convergence rate. By solving the dual problem using the semismooth Newton method, the scaled proximal mapping is obtained efficiently, contributing to the effectiveness and computational efficiency of the APSSN algorithm. Both theoretical analysis and empirical study support the effectiveness of APSSN for composite function optimization problems.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2021)
Article
Mathematics
Dawan Chumpungam, Panitarn Sarnmeta, Suthep Suantai
Summary: In this paper, a new line search technique is introduced and utilized to develop an accelerated forward-backward algorithm for solving convex minimization problems with two convex functions in a real Hilbert space. The algorithm's weak convergence is established without requiring the Lipschitz assumption on the objective function's gradient. Furthermore, its performance is analyzed through its application to classification problems on various data sets and comparison with other line search algorithms. Based on experimental results, the proposed algorithm outperforms other line search algorithms.
Article
Engineering, Civil
Zhong-kai Feng, Wen-jing Niu, Zheng-yang Tang, Yang Xu, Hai-rong Zhang
Summary: A novel evolutionary artificial intelligence model is developed for multiple scales nonstationary hydrological time series prediction, utilizing the cooperation search algorithm to optimize the ELM model's input-hidden weights and biases. Experimental results show that the proposed method outperforms the traditional ELM method in terms of performance evaluation indexes, particularly with significant improvements in both RMSE and MAPE during testing phase, supporting decision-making in water resource system.
JOURNAL OF HYDROLOGY
(2021)
Article
Computer Science, Artificial Intelligence
Tao Sun, Dongsheng Li
Summary: This paper considers a class of nonconvex regularized optimization problems and proposes an accelerated algorithm. The algorithm inherits the advantages of decentralized algorithms and is proven to converge.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2022)
Article
Computer Science, Information Systems
R. Thenmozhi, Abdul Wahid Nasir, Vijaya Krishna Sonthi, T. Avudaiappan, Seifedine Kadry, Kuntha Pin, Yunyoung Nam
Summary: This article introduces a new Sparrow Search Algorithm with Doppler Effect (SSA-DE) for Node Localization in Wireless Networks. By incorporating bio-inspired algorithms and the Doppler Effect, the performance and efficiency of node localization are improved.
CMC-COMPUTERS MATERIALS & CONTINUA
(2022)
Article
Computer Science, Software Engineering
Amirhossein Zolfagharian, Manel Abdellatif, Lionel C. Briand, Mojtaba Bagherzadeh, S. Ramesh
Summary: This paper proposes a search-based testing approach for reinforcement learning agents to test their policies by effectively searching for failing executions. The approach relies on machine learning models and a genetic algorithm to narrow down the search for detecting more faults related to the agent's policy. The authors also investigate how to extract rules from the search results to understand the conditions under which the agent fails and assess the deployment risks.
IEEE TRANSACTIONS ON SOFTWARE ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Ruobin Gao, Jiahui Liu, Qin Zhou, Okan Duru, Kum Fai Yuen
Summary: Asset prices are crucial for the financial survival and profitability of ship-owning firms. This paper proposes an improved intelligent model search engine that uses asynchronous time-lag selection to predict ship prices and achieve more accurate results.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Haishan Ye, Chaoyang He, Xiangyu Chang
Summary: This article proposes a novel distributed second-order optimization algorithm called accelerated distributed approximate Newton (ADAN), which can effectively reduce computation costs and achieve both communication and computation efficiencies. Compared with other algorithms, ADAN is based on the inexact Newton theory, enabling efficient handling of expensive subproblems and improving optimization performance.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Yang-yang Liu, Jian-wei Liu
Summary: The most essential purpose of the machine learning field is to minimize the difference between the true value and the predicted value. Researchers have proposed a loss function as a learning criterion, which can be connected with optimization learning. Gradient descent is commonly used in machine learning, but it faces the problem of divergence. To solve this problem, a homotopy analysis method is proposed, which effectively controls convergence without divergence, and can reduce the number of iterations.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Construction & Building Technology
Fatemeh Nejati, Wahidullah Omer Zoy, Nayer Tahoori, Pardayev Abdunabi Xalikovich, Mohammad Amin Sharifian, Moncef L. Nehdi
Summary: This research proposes a novel machine learning tool for simulating building thermal load by using a symbiotic organism search algorithm applied to an artificial neural network. The results show that the SOS-ANN hybrid is a strong predictor and is highly recommended for early estimation of heating load in buildings.
Article
Construction & Building Technology
Ligang Shi, Jinghan Qiu, Ruinan Zhang, Yuqing Li, Zhaojing Yang, Xinzhu Qi, Lulu Tao, Siying Li, Weiming Liu
Summary: This study aimed to establish a computational method for assessing visual comfort from the human-centric perspective in a university gymnasium. Through mutual authentication between questionnaire and physiological indices and luminance, the correlation between various luminance levels and human perception was quantified using machine learning. Based on the artificial neural networks, the most optimized visual comfort assessment model was determined, with a correlation coefficient between luminance, synthetical visual evaluation (SVE), and physiological indicators ranging from 85% to 90%. According to the genetic algorithm, a comfortable visualization requires an average luminance of 55-135 cd/m2 for the entire field of view (Lfov), 82-375 cd/m2 for the target area (Lt), and 960-1950 cd/m2 for the window area (Lw).
JOURNAL OF BUILDING ENGINEERING
(2023)
Article
Multidisciplinary Sciences
Juan Wei, Chao Sun, Xue-jing Zhang, Er-jie Wang, Deify Law
Summary: A subsection interpolation method based on curve curvature threshold is proposed to solve the incompatible problem of machining accuracy and efficiency in parametric curve machining. The method calculates the curvature threshold based on geometric and kinematic constraints, and determines the interpolation key points and their velocities using the threshold points and start and end points of the curve. The adaptive Simpson method is used to calculate the arc length of each subsegment. Real-time interpolation is achieved using the parametric modified second-order Runge-Kutta method, which significantly improves interpolation accuracy and reduces interpolation time. Numerical cases show that the proposed method can smooth the overall interpolation speed, reduce speed fluctuation, and improve real-time performance.
SCIENTIFIC REPORTS
(2022)
Article
Multidisciplinary Sciences
Bing Hao, Jianshuo Zhao, He Du, Qi Wang, Qi Yuan, Shuo Zhao
Summary: The search algorithm plays a crucial role in the robot's motion planning and task completion. A fusion algorithm combining Flower Pollination algorithm and Q-learning is proposed to solve search tasks in complex environments. Improving the accuracy and efficiency of the search and rescue robot path search, the algorithm uses an improved grid map and combines Q-learning and Flower Pollination algorithm for initialization of Q-table. It also introduces a combination of static and dynamic reward function to better accommodate different situations encountered by the robot during the search process. Experiment results demonstrate the success of the improved grid map and the effectiveness of the FIQL algorithm, which outperforms other algorithms in terms of reducing iterations, improving adaptability to complex environments, and minimizing computational effort.
Article
Computer Science, Artificial Intelligence
Liang-Ching Chen, Kuei-Hu Chang
Summary: This study explores the application value and method of using a corpus to tackle novel diseases like COVID-19. A new approach combining corpus software and H-index algorithm to handle word-ranking issues is proposed. Empirical evidence and comparative results indicate that this method is more accurate in handling word-ranking issues.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2021)