检索结果-内蒙古大学图书馆

AceGPT, Localizing Large Language Models in Arabic

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Huang, Huang Yu, Fei Zhu, Jianqing Sun, Xuening Cheng, Hao Song, Dingjie Chen, Zhihong Alharthi, Abdulmohsen An, Bang He, Juncai Liu, Ziche Zhang, Zhiyi Chen, Junying Li, Jianquan Wang, Benyou Zhang, Lian Sun, Ruoyu Wan, Xiang Li, Haizhou Xu, Jinchao Shenzhen International Center for Industrial and Applied Mathematics Shenzhen Research Institute of Big Data China The Chinese University of Hong Kong Shenzhen China King Abdullah University of Science and Technology Thuwal Saudi Arabia Shenzhen Research Institute of Big Data Shenzhen China

This paper is devoted to the development of a localized Large Language Model (LLM) specifically for Arabic, a language imbued with unique cultural characteristics inadequately addressed by current mainstream models. Significant concerns emerge when addressing cultural sensitivity and local values. To address this, the paper proposes a comprehensive solution that includes further pre-training with Arabic texts, Supervised Fine-Tuning (SFT) utilizing native Arabic instructions, and GPT-4 responses in Arabic, alongside Reinforcement Learning with AI Feedback (RLAIF) employing a reward model attuned to local culture and values. The goal is to cultivate culturally cognizant and value-aligned Arabic LLMs capable of accommodating the diverse, application-specific needs of Arabic-speaking communities. Comprehensive evaluations reveal that the resulting model, dubbed ‘AceGPT’, sets the state-of-the-art standard for open Arabic LLMs across various benchmarks. Codes, data, and models are in https://***/FreedomIntelligence/AceGPT. © 2023, CC BY-NC-SA.

关键词： Reinforcement learning

Constraints on Dark Matter-Neutrino Interaction from 21-cm Cosmology and Forecasts on SKA1-Low

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Dey, Antara Paul, Arnab Pal, Supratik Physics and Applied Mathematics Unit Indian Statistical Institute Kolkata700108 India School of Physical Sciences Indian Association for the Cultivation of Science Kolkata700032 India Technology Innovation Hub on Data Science Big Data Analytics and Data Curation Indian Statistical Institute Kolkata700108 India

In this article, we have done a thorough investigation of the possible effects of interaction between dark matter (DM) and neutrinos on reionization history. We have constrained the interaction strength using 21 cm Cosmology and found out possible deviations from standard, non-interacting ΛCDM scenario. Comparing the results with the existing constraints from present cosmological observations reveals that 21 cm observations are more competent to constrain the interaction strength by a few orders of magnitude. We have also searched for prospects of detecting any such interaction in the upcoming 21 cm mission SKA1-Low by doing a forecast analysis and error estimation. Copyright © 2022, The Authors. All rights reserved.

关键词： Dark Matter

Price Prediction of Seasonal Items Using Time Series Analysis

学校读者我要写书评

暂无评论

Computer Systems science & Engineering 2023年第7期46卷 445-460页

作者： Ahmed Salah Mahmoud Bekhit Esraa Eldesouky Ahmed Ali Ahmed Fathalla Faculty of Computers and Informatics Zagazig UniversitySharkeya44523Egypt Information Technology Department the University of Technology and Applied SciencesIbriOman School of Electrical and Data Engineering University of Technology SydneySydney2007Australia College of Computer Engineering and Sciences Prince Sattam Bin Abdulaziz UniversityAl-Kharj11942Saudi Arabia Computer Science Department Faculty of Computers and InformaticsSuez Canal UniversityIsmailia41522Egypt Higher Future Institute for Specialized Technological Studies Cairo3044Egypt Department of Mathematics Faculty of ScienceSuez Canal UniversityIsmailia41522Egypt

The price prediction task is a well-studied problem due to its impact on the business *** are several research studies that have been conducted to predict the future price of items by capturing the patterns of price change,but there is very limited work to study the price prediction of seasonal goods(e.g.,Christmas gifts).Seasonal items’prices have different patterns than normal items;this can be linked to the offers and discounted prices of seasonal *** lack of research studies motivates the current work to investigate the problem of seasonal items’prices as a time series *** proposed utilizing two different approaches to address this problem,namely,1)machine learning(ML)-based models and 2)deep learning(DL)-based ***,this research tuned a set of well-known predictive models on a real-life *** models are ensemble learning-based models,random forest,Ridge,Lasso,and Linear ***,two new DL architectures based on gated recurrent unit(GRU)and long short-term memory(LSTM)models are ***,the performance of the utilized ensemble learning and classic ML models are compared against the proposed two DL architectures on different accuracy metrics,where the evaluation includes both numerical and visual comparisons of the examined *** obtained results show that the ensemble learning models outperformed the classic machine learning-based models(e.g.,linear regression and random forest)and the DL-based models.

关键词： Deep learning price prediction seasonal goods time series analysis

Modeling and Flux Reconstruction of Unstructured Geometric Reactor

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Xiao, Bowen Zheng, Youqi Wei, Linfang Institute of Applied Physics and Computational Mathematics Beijing100094 China School of Nuclear Science and Technology Xi’an Jiaotong University Shaanxi Xi’an710094 China School of Optoelectronic Engineering Xidian University Shaanxi Xi’an710071 China

Innovative reactors are typically designed with complex geometric structures, which put forward higher requirements for deterministic reactor core calculation. In the complex geometric reactor core calculation, the LAVENDER code in the SARAX code system was used, which is based on the three-dimensional neutron transport nodal method. A modeling method based on constructive solid geometry was applied to generate arbitrary triangular-z meshes. To meet the requirements of coupling core physics calculation with thermal and other physics fields, a flux reconstruction method based on the statistics of biharmonic spline interpolation of Green’s function has been proposed, which was used to reconstruct the flux of specified region. Three reactors with different types were calculated to verify the accuracy of the flux reconstruction method. Numerical results show that the fluxes of specified region calculated by the proposed flux reconstruction method in the SARAX code were in good agreement with the direct full-core Monte-Carlo results. © 2024, The Authors. All rights reserved.

关键词： Interpolation

Influence of the two-temperature effect on ionization potential depression in hot dense plasma

学校读者我要写书评

暂无评论

Physical Review E 2025年第3期111卷 035208-035208页

作者： Qingbo Luo Xin Liang Chengliang Lin Xinlian Zhang Jianpeng Liu Cheng Gao Yong Hou Jianmin Yuan College of Science National University of Defense Technology Changsha Hunan 410073 People's Republic of China Hunan Key Laboratory of Extreme Matter and Applications National University of Defense Technology Changsha Hunan 410073 People's Republic of China National Key Laboratory of Computational Physics Institute of Applied Physics and Computational Mathematics Beijing 100094 People's Republic of China Institute of Atomic and Molecular Physics Jilin University Changchun Jilin 130012 People's Republic of China

In hot dense plasma, the interaction between charged particles leads to the ionization potential depression (IPD), which further affects the physical properties of plasma, such as opacity and equation of state. The experiment of IPD of solid-density Al plasma has indicated that present theoretical models cannot give reasonable descriptions of the IPD in hot dense plasma. So, reasonable theoretical methods are needed to describe the effects of hot dense environments on IPD, and the process of generating hot dense plasmas through the interaction between ultrashort laser pulse and solid-density matter also needs to be carefully considered. In the manuscript, two kinds of temperatures for ions and electrons are considered when we compute the ionization potential depression in hot dense plasma. And, the influences of the hot dense environments are included in the electronic structure calculations by using the modified flexible atomic code (FAC), which has included the screening of free electrons and the correlation of ions by correlation functions obtained from the hyper-netted chain (HNC) approximation. A self-consistent-field method is used to calculate the electronic structures. Based on the calculations, the IPD is obtained through the two-step model. Considering the interaction of the femtosecond laser on the solid-density Al plasma of Ciricosta's experiment, we use the two-temperature model to calculate the IPD in nonlocal thermodynamic equilibrium, and the theoretical results are in good agreement with the experimental results. In addition, we also calculated the electron collision ionization cross section and compared it with the results from other models.

关键词： Ionization potential

SWAP Attack: Stealthy Side-Channel Attack on Multi-Tenant Quantum Cloud System

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Lee, Wei Jie Bryan Wang, Siyi Dutta, Suman Maouaki, Walid El Chattopadhyay, Anupam College of Computing and Data Science Nanyang Technological University Singapore Applied Statistics Unit Indian Statistical Institute Kolkata India Quantum Physics and Spintronics Team LPMC Faculty of Sciences Ben M’Sik Hassan II University of Casablanca Morocco

The rapid advancement of quantum computing has spurred widespread adoption, with cloud-based quantum devices gaining traction in academia and industry. This shift raises critical concerns about the privacy and security of computations on shared, multi-tenant quantum platforms accessed remotely. In several recent works, crosstalk has been exploited on shared quantum devices, enabling adversaries to interfere with the victim circuits within a neighborhood. Though illuminating, these studies left several unresolved questions regarding the fundamental cause of crosstalk, effective countermeasures, and replicability across different circuits. In this paper, we revisit the crosstalk effect, tracing its origins to the SWAP path between various qubits in a circuit and demonstrating its effectiveness even over long distances. The results significantly improve our understanding of the phenomenon beyond prior works. Our proposed SWAP-based side-channel attack can be executed in both active and passive modes, as verified on real shared IBM quantum devices. In the active attack, an adversary executing a single CNOT gate can perturb victim circuits running Grover’s Algorithm from afar, reducing the expected output accuracy through 81.62% by strategically positioning qubits. Moreover, this phenomenon can be clearly explained through modeling, where specific qubits in a platform can be identified as more susceptible to an attacker. The passive attack, leveraging a stealthy circuit as small as 6.25% of the victim’s, achieves 100% accuracy in predicting the circuit size of the victim, running Simon’s Algorithm. These findings challenge the existing defense strategy of maximizing topological distance between circuits, demonstrating that attackers can still extract sensitive information or manipulate results remotely. Our work highlights the need for more robust security measures to safeguard quantum computations against such emerging threats. © 2025, CC BY.

关键词： Qubits

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Jing, Bing-Yi Li, Ting Wang, Jiangzhou Wang, Ya Department of Statistics and Data Science Southern University of Science and Technology Shenzhen518055 China Department of Applied Mathematics The Hong Kong Polytechnic University Hong Kong999077 Hong Kong School of Mathematical Sciences Institute of Statistical Sciences Shenzhen University Shenzhen518060 China

There has been extensive research on community detection in directed and bipartite networks. However, these studies often fail to consider the popularity of nodes in different communities, which is a common phenomenon in real-world networks. To address this issue, we propose a new probabilistic framework called the Two-Way Node Popularity Model (TNPM). The TNPM also accommodates edges from different distributions within a general sub-Gaussian family. We introduce the Delete-One-Method (DOM) for model fitting and community structure identification, and provide a comprehensive theoretical analysis with novel technical skills dealing with sub-Gaussian generalization. Additionally, we propose the Two-Stage Divided Cosine Algorithm (TSDC) to handle large-scale networks more efficiently. Our proposed methods offer multi-folded advantages in terms of estimation accuracy and computational efficiency, as demonstrated through extensive numerical studies. We apply our methods to two real-world applications, uncovering interesting findings. Copyright © 2024, The Authors. All rights reserved.

关键词： Computational efficiency

Powerful laser-produced quasi-half-cycle THz pulses

学校读者我要写书评

暂无评论

Physical Review E 2022年第3期106卷 035201-035201页

作者： A. S. Kuratov A. V. Brantov V. F. Kovalev V. Yu. Bychenkov P. N. Lebedev Physics Institute Russian Academy of Science Leninskii Prospect 53 Moscow 119991 Russia Center for Fundamental and Applied Research Dukhov Research Institute of Automatics (VNIIA) Moscow 127055 Russia Keldysh Institute of Applied Mathematics Russian Academy of Sciences Moscow 125047 Russia

The Maxwell equations-based 3D-analytical solution for the terahertz (THz) half-cycle electromagnetic wave transition radiation pulse has been found. This solution describes generation and propagation of transition radiation into free space from laser-produced relativistic electron bunch which crosses a target-vacuum interface as a result of ultrashort laser pulse interaction with a thin high-conductivity target. The analytical solution found complements the theory of laser initiated transition radiation. It describes the THz wave half-cycle pulse at the arbitrary distance from a target surface including near-field zone rather than its standard far-field characterization. The analytical research has also been supplemented with the 3D simulations using the finite-difference time-domain method, which makes it possible for description of much wider spatial domain as compared to that from the particle-in-cell approach. The presented result sheds light fundamentally on the interference of the electron bunch field and the generated THz field of broadband transition radiation from laser-plasma interaction. The latter is studied for a long time in the experiments with solid density plasma and the theory developed may inspire to targeted measurements and investigations of unique super intense half-cycle THz radiation waves near the laser target.

关键词： Laser-plasma interactions Terahertz generation in plasmas

Fast-Convergent and Communication-Alleviated Heterogeneous Hierarchical Federated Learning in Autonomous Driving

学校读者我要写书评

暂无评论

IEEE Transactions on Intelligent Transportation Systems 2025年

作者： Kou, Wei-Bin Lin, Qingfeng Tang, Ming Ye, Rongguang Wang, Shuai Zhu, Guangxu Wu, Yik-Chung The University of Hong Kong Department of Electrical and Electronic Engineering Hong Kong Hong Kong Southern University of Science and Technology Department of Computer Science and Engineering Shenzhen518055 China Chinese Academy of Sciences Shenzhen Institute of Advanced Technology Shenzhen518055 China Shenzhen International Center for Industrial and Applied Mathematics Shenzhen Research Institute of Big Data Shenzhen518172 China

Street Scene Semantic Understanding (denoted as TriSU) is a complex task for autonomous driving (AD). However, inference model trained from data in a particular geographical region faces poor generalization when applied in other regions due to inter-city data domain-shift. Hierarchical Federated Learning (HFL) offers a potential solution for improving TriSU model generalization by collaborative privacy-preserving training over distributed datasets from different cities. Unfortunately, it suffers from slow convergence because the data from different cities are with disparate statistical properties. Going beyond existing HFL methods, we propose a Gaussian heterogeneous HFL algorithm (FedGau) to address inter-city data heterogeneity so that convergence can be accelerated. In the proposed FedGau algorithm, both single RGB image and RGB dataset are modelled as Gaussian distributions for aggregation weight design. This approach not only differentiates each RGB image by respective statistical distribution, but also exploits the statistics of dataset from each city in addition to the conventionally considered data volume. With the proposed approach, the convergence is accelerated by 35.5%-40.6% compared to existing state-of-the-art (SOTA) HFL methods. On the other hand, to reduce the involved communication resource, we further introduce a novel performance-aware adaptive resource scheduling (AdapRS) policy. Unlike the traditional static resource scheduling policy that exchanges a fixed number of models between two adjacent aggregations, AdapRS adjusts the number of model aggregation at different levels of HFL so that unnecessary communications are minimized. Extensive experiments demonstrate that AdapRS saves 29.65% communication overhead compared to conventional static resource scheduling policy while maintaining almost the same performance. © 2000-2011 IEEE.

关键词： Gaussian distribution