Article
Mathematics, Applied
Guangkui Xu, Gaojun Luo, Xiwang Cao
Summary: In this paper, we conducted a further study on the permutation behavior of polynomials of a special form by considering the number of solutions of certain equations over finite fields. We presented four classes of permutation polynomials of the form (x(2m) + x + delta)(s) + x over F-22m, and provided necessary and sufficient conditions for these polynomials to permute F-22m. Furthermore, we introduced several classes of permutation polynomials of the form (x(pm) - x + delta)(s) + x over F-p2m of odd characteristic, some of which can provide complete permutation polynomials of this form over F-32m.
FINITE FIELDS AND THEIR APPLICATIONS
(2022)
Article
Mathematics, Applied
Xiang-dong Hou, Vincenzo Pallozzi Lavorante
Summary: A general construction is provided to generate permutation polynomials of Fq2 with certain properties, including previously unknown polynomials. This construction is particularly suitable for permutation binomials and trinomials.
FINITE FIELDS AND THEIR APPLICATIONS
(2023)
Article
Biochemistry & Molecular Biology
Chaochao Yan, Peilin Zhao, Chan Lu, Yang Yu, Junzhou Huang
Summary: The main goal of retrosynthesis is to break down desired molecules into available building blocks. Existing template-based methods are limited by a fixed template selection pattern and lack the ability to discover new reactions. To overcome this limitation, we propose an innovative retrosynthesis prediction framework that can generate new templates beyond the training templates. We also introduce an effective reactant candidate scoring model that captures atom-level transformations and outperforms previous methods. Experimental results demonstrate the ability of our method to generate novel templates for reactions not covered by training templates.
Article
Computer Science, Artificial Intelligence
Pablo Morala, Jenny Alexandra Cifuentes, Rosa E. Lillo, Inaki Ucar
Summary: This article discusses a mathematical framework relating neural networks and polynomial regression, and proposes a new method for their relationship. Through experimental validation, it is shown that this method can produce polynomials that correctly approximate data when learning from polynomial generated data.
Article
Chemistry, Physical
Carly E. LaGrotta, Qinghui Meng, Lei Lei, Mark C. Barbet, Zekai Hong, Michael P. Burke
Summary: Recent high-level theoretical calculations suggest a different temperature dependence for HO2 + HO2 compared to previous experiments. By analyzing the theoretical and experimental data, we found that a different interpretation of the experimental data can reconcile the discrepancies, showing consistency between theory and experiment. The presence of an HOOOOH intermediate, identified by recent theory, but not considered in earlier interpretations, may have contributed to the scatter in data.
JOURNAL OF PHYSICAL CHEMISTRY A
(2023)
Article
Chemistry, Analytical
Yifan Xiao, Zhixin Guo, Peter Veelaert, Wilfried Philips
Summary: This paper proposes a unified and flexible framework for general image fusion tasks, which can handle multiple fusion tasks and use symmetrical functions to extract salient features from input images for fusion. Continual learning based on Elastic Weight Consolidation (EWC) is applied to handle different fusion tasks.
Article
Engineering, Electrical & Electronic
Xiao Yu, Bing Xia, Shuxin Yang, Hongshen Yin, Yajie Wang, Xiaowen Liu
Summary: In recent years, the deep learning-based fault diagnosis methods for rotating mechanical equipment have attracted attention. However, due to differences in data feature distributions under varying working conditions, these models cannot provide satisfactory fault prediction performance in such scenarios. To address this, this paper proposes a domain adversarial-based rolling bearing fault transfer diagnosis model EMBRNDNMD. The model uses an EEMD-based time-frequency feature graph construction method to extract time-frequency information and a multi-branch ResNet structure to extract deep features representing the bearing state. Additionally, an adversarial network module and MK-MMD distribution difference evaluation method are introduced to optimize the model and reduce the probability distribution difference between source and target domains, thus improving the accuracy of EMBRNDNMD in the target domain.
JOURNAL OF SENSORS
(2022)
Article
Astronomy & Astrophysics
Yusuke Namekawa, Kouji Kashiwa, Akira Ohnishi, Hayato Takase
Summary: This paper investigates the efficiency of a gauge invariant input for neural network path optimization. By using a gauge invariant input, such as a plaquette, the sign problem is successfully tamed in a two-dimensional U(1) gauge theory with a complex coupling. This opens up the possibility of applying path optimization to complex gauge theories, including quantum chromodynamics, in a realistic setup.
Article
Computer Science, Interdisciplinary Applications
Danyao Wu, Pingzhi Yuan
Summary: In this paper, the permutation behavior of a class of polynomials over finite fields is investigated using the AGW criterion. The results are generalized by introducing special exponents s.
APPLICABLE ALGEBRA IN ENGINEERING COMMUNICATION AND COMPUTING
(2022)
Article
Biochemical Research Methods
Cheng Wang, Chuang Yuan, Yahui Wang, Ranran Chen, Yuying Shi, Tao Zhang, Fuzhong Xue, Gary J. Patti, Leyi Wei, Qingzhen Hou
Summary: This study developed a VGAE-based framework to predict Metabolite-Protein Interaction (MPI) in genome-scale enzymatic reaction networks. The predictor achieved the best performance compared to other methods by incorporating molecular features of metabolites and proteins as well as neighboring information in the MPI networks. Furthermore, the framework was used to reconstruct disease-specific MPI networks and identify novel enzymatic reaction links.
BRIEFINGS IN BIOINFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Liang Ding, Longyue Wang, Siyou Liu
Summary: This article introduces a recursive graph syntax encoder called RGSE, which can simultaneously model syntactic dependencies and sequential information in neural machine translation tasks. The study shows that models equipped with RGSE outperform existing models on multiple strong syntax-aware benchmarks.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2023)
Article
Chemistry, Multidisciplinary
Yong-Seok Lee, Dong-Won Jang
Summary: A neural network method for self-tuning PID controllers was proposed, which significantly reduces tuning attempts, achieves a performance efficiency of 92.9%, and is also applicable for traditional PID controllers.
APPLIED SCIENCES-BASEL
(2021)
Review
Computer Science, Information Systems
Zengchen Yu, Ke Wang, Zhibo Wan, Shuxuan Xie, Zhihan Lv
Summary: Due to its automatic feature learning ability and high performance, deep learning has gradually become the mainstream of artificial intelligence in recent years, playing a role in many fields. This paper introduces several deep learning algorithms such as Artificial Neural Network (NN), FM-Deep Learning, Convolutional NN and Recurrent NN, and explains their theory, development history, and applications in disease prediction. The paper also analyzes the current defects in the disease prediction field and provides some current solutions. Furthermore, it discusses two major trends in the future disease prediction and medical field - integrating Digital Twins and promoting precision medicine.
CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS
(2023)
Article
Remote Sensing
Xinyu Liu, Xiaoguang Di, Junde Wu, Jiehao Huang
Summary: The proposed convolutional neural network architecture for object detection in high-resolution remote-sensing images is anchor-free and outperforms previous methods in terms of detecting accuracy without adding extra trainable parameters.
INTERNATIONAL JOURNAL OF REMOTE SENSING
(2021)
Article
Mathematics
Nhat-Duc Hoang
Summary: This paper conducts a comparative study on the predictive capability of machine learning models for estimating the compressive strength of self-compacting concrete. The results indicate that deep neural network regression (DNNR) is the best model, followed by extreme gradient boosting machine (XGBoost). These models show great potential in modeling the compressive strength of self-compacting concrete.
Article
Chemistry, Organic
Han Luo, Yumeng Li, Yuan Zhang, Qixing Lu, Qiaoyu An, Mingchuan Xu, Shanshan Li, Jun Li, Baosheng Li
Summary: In this study, a nonclassic mechanism for the concerted SNAr reaction was developed and applied for the synthesis of biologically important compounds. This reaction offers a convenient and efficient approach to access the desired products.
JOURNAL OF ORGANIC CHEMISTRY
(2022)
Article
Physics, Atomic, Molecular & Chemical
Meng Zhang, Yongfa Zhu, Jun Li
Summary: This study investigates the dynamics of the hydrogen abstraction reaction between methanol and fluorine atoms, providing valuable insights into the importance of this reaction and the vibrational state distribution of the resulting products.
CHINESE JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Jie Qin, Yang Liu, Jun Li
Summary: In this study, an accurate full-dimensional potential energy surface (PES) was developed, and the non-minimum energy path (MEP) dynamics of the reaction between OH- and CH3F were revealed through dynamic simulations, uncovering new proton exchange and proton abstraction product channels, providing new insights into the mechanism of the reaction.
JOURNAL OF CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Xiaohu Xu, Jun Li
Summary: Chemical reaction dynamics requires the collaboration of both experiment and theory. With the help of machine learning, an accurate potential energy surface (PES) for the reaction between Cl + SiH4 has been developed. This PES successfully describes both the hydrogen abstraction channel and the hydrogen substitution channel, revealing detailed dynamics of the latter. Theoretical and experimental results suggest that the substitution channel mainly occurs through the typical S(N)2 inversion mechanism.
JOURNAL OF PHYSICAL CHEMISTRY A
(2022)
Article
Chemistry, Physical
Xiaomin Lin, Wei Wang, Bing He, Jun Li, Qiantao Wang, Jianyi Ma
Summary: In this study, a simplified and accurate general AMOEBA polarizable force field called Combustion-AMOEBA or cAMOEBA is reported. The cAMOEBA force field eliminates permanent atomic dipoles and quadrupoles, retains explicit polarization, and defines general atom types for different molecular species. It avoids the tedious parameterization process for new molecules in the original AMOEBA force field, enabling efficient high-throughput calculations for a large number of molecules. The cAMOEBA force field shows good consistency with the original AMOEBA force field, and its parameters can be used to develop a high-quality transport property database for combustion modeling.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Chemistry, Physical
Chun Tao, Jiawei Yang, Qizhen Hong, Quanhua Sun, Jun Li
Summary: The energy transfer, dissociations, and chemical reactions between O2 and N2 are crucial in aircraft re-entry and various atmospheric, combustion, and plasma processes. In this study, a high-precision potential energy surface (PES) for the O2 and N2 system was developed based on 55,000 data points. The performances of different fitting methods and the resulting PESs were compared through various scans, properties of stationary points, and dynamics simulations. Suggestions for improving the PES of N2O2 were also discussed for future high-temperature calculations and simulations.
JOURNAL OF PHYSICAL CHEMISTRY A
(2023)
Article
Chemistry, Physical
Qi Zhang, Jun Li
Summary: This study reveals the microsolvation process of NaCl(H2O) and NaCl(H2O)(-) salts in water through comprehensive calculations and analysis. The structural rearrangements of the neutral NaCl(H2O) and NaCl(H2O)(-) anions are mainly caused by breaking and forming of hydrogen bonds and the enhancement and weakening of interactions between Na and O atoms. The distributions of anion complexes and the photoelectron spectra of the anions are computed and analyzed, showing good agreement with experimental results.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2023)
Article
Chemistry, Physical
Jia Li, Yang Liu, Hua Guo, Jun Li
Summary: In this study, a high-precision full-dimensional potential energy surface (PES) was developed using a combination of two ab initio methods. The reaction dynamics and mode specificity of the hydrogen exchange channel between H2 and H-2 were investigated. It was found that different vibrational modes can promote the reaction, with the H-H stretching mode having the strongest promotion effect. The Sudden Vector Projection (SVP) model was applied to predict mode specificity effects and explain the product energy partitioning. The hydrogen exchange channel was dominated by sideways scattering.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Kaisheng Song, Hongwei Song, Jun Li
Summary: This study develops a globally accurate potential energy surface and investigates the inhibitory effect of exciting the rotational mode of H-2 on the H-2 + NH2- -> NH3 + H- reaction. The findings are consistent with experimental results and provide insights into the mechanism of ion-molecule reactions.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2022)
Article
Chemistry, Physical
Yang Liu, Jun Li
Summary: The study introduces a neural network-based Delta-machine learning approach for efficiently constructing full-dimensional accurate potential energy surfaces of complex reactions. The flexibility of the neural network is utilized to efficiently sample points from the low-level data set and successfully elevate the newly fitted potential energy surface to a high-quality level.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2022)