检索结果-内蒙古大学图书馆

AuGQ: Augmented quantization granularity to overcome accuracy degradation for sub-byte quantized deep neural networks

引用

APPLIED INTELLIGENCE 2025年第7期55卷 1-20页

作者： Mujtaba, Ahmed Lee, Wai Kong Ko, Byoung Chul Chang, Hyung Jin Hwang, Seong Oun Gachon Univ Dept IT Convergence Engn Seongnam 13120 South Korea Gachon Univ Dept Comp Engn Seongnam 13120 South Korea Keimyung Univ KMU Dept Comp Engn Daegu South Korea Univ Birmingham Sch Comp Sci Birmingham England

Deployment of neural networks on IoT devices unleashes the potential for various innovative applications, but the sheer size and computation of many deep learning (DL) networks prevented its widespread. Quantization mitigates this issue by reducing model precision, enabling deployment on resource-constrained edge devices. However, at extremely low bit-widths, such as 2-bit and 4-bit, the aggressive compression leads to significant accuracy degradation due to the reduced representational capacity of the neural network. A critical aspect of effective quantization is identifying the range of real values (FP32) that impact model accuracy. To address accuracy loss at sub-byte levels, we introduce Augmented Quantization (AuGQ), a novel granularity technique tailored for low bit-width quantization. AuGQ segments the range of real-valued (FP32) weight and activation distributions into small uniform intervals, applying affine quantization in each interval to enhance accuracy. We evaluated AuGQ using both post-training quantization (PTQ) and quantization-aware training (QAT) methods, achieving accuracy levels comparable to full precision (32-bit) DL networks. Our findings demonstrate that AuGQ is agnostic to the training pipeline and batch normalization folding, distinguishing it from conventional quantization techniques. Furthermore, when integrated into state-of-the-art PTQ algorithms, AuGQ necessitates only 64 training samples for fine-tuning which is 16x fewer than traditional methods. This reduction facilitates the application of high-accuracy quantization at sub-byte bit-widths, making it suitable for practical IoT deployments and enhancing computational efficiency on edge devices.

关键词： Deep Neural Networks (DNNs) Edge computing Internet-of-Things (IoT) optimization algorithms Quantization of neural networks TinyML

来源：评论

学校读者我要写书评

暂无评论

A comprehensive survey of fractional gradient descent methods and their convergence analysis

引用

CHAOS SOLITONS & FRACTALS 2025年 194卷

作者： Elnady, Sroor M. El-Beltagy, Mohamed Radwan, Ahmed G. Fouda, Mohammed E. Cairo Univ Fac Engn Engn Math & Phys Dept Giza Egypt Nile Univ Sch Engn & Appl Sci Giza Egypt Compumacy Artificial Intelligence Solut Cairo Egypt

Fractional Gradient Descent (FGD) methods extend classical optimization algorithms by integrating fractional calculus, leading to notable improvements in convergence speed, stability, and accuracy. However, recent studies indicate that engineering challenges-such as tensor-based differentiation in deep neural networks-remain partially unresolved, prompting further investigation into the scalability and computational feasibility of FGD. This paper provides a comprehensive review of recent advancements in FGD techniques, focusing on their approximation methods and convergence properties. These methods are systematically categorized based on their strategies to overcome convergence challenges inherent in fractional-order calculations, such as nonlocality and long-memory effects. Key techniques examined include modified fractional-order gradients designed to avoid singularities and ensure convergence to the true extremum. Adaptive step-size strategies and variable fractional-order schemes are analyzed, balancing rapid convergence with precise parameter estimation. Additionally, the application of truncation methods is explored to mitigate oscillatory behavior associated with fractional derivatives. By synthesizing convergence analyses from multiple studies, insights are offered into the theoretical foundations of these methods, including proofs of linear convergence. Ultimately, this paper highlights the effectiveness of various FGD approaches in accelerating convergence and enhancing stability. While also acknowledging significant gaps in practical implementations for large-scale engineering tasks, including deep learning. The presented review serves as a resource for researchers and practitioners in the selection of appropriate FGD techniques for different optimization problems.

关键词： Fractional calculus optimization algorithms Gradient descent Variable fractional order

来源：评论

学校读者我要写书评

暂无评论

Machine learning approaches for predicting mechanical properties of steel-fiber-reinforced concrete

引用

MATERIALS TODAY COMMUNICATIONS 2025年 45卷

作者： Huang, Huan Lei, Long Xu, Guoxin Cao, Shuangli Ren, Xiping Shaanxi Univ Int Trade & Commerce Xian 712046 Shaanxi Peoples R China Hanjiang Weihe River Water Divers Project Construc Xian 710024 Shaanxi Peoples R China

Recent advancements in machine learning enable accurate prediction of concrete's mechanical properties, offering efficiency in data analysis and reducing material waste in civil engineering. What sets this study apart from the literature is the simultaneous prediction of Compressive, Tensile, and Flexural Strength for reinforced concrete (SFRC). This integrated prediction could streamline the material selection process, enhance the durability of concrete structures, and improve performance in real-world applications such as pavements, bridges, and industrial floors. Unique variables like fiber length and content are among the input variables for predictions, which may lead to more cost-effective structural solutions. In this study, base models of the Ridge Regression and Quantile Regression were cross-validated, and the best folds of the compiled dataset for training and testing models were selected. The Sled Dog Optimizer, as a recently developed strong optimization algorithm, is utilized for fine-tuning hyperparameters of the models and enhancing prediction performance. Also, the Bayesian model combination method was employed to develop a hybrid ensemble model for reliably predicting concrete strengths by combining the capabilities of both hybrid models. Such hybrid and ensemble models are rare in the literature in this field. Also, SHAP-based explainable AI utilized for interpreting importance of variables in predictions. The hybrid ensemble model performed best in predicting Compressive Strength, achieving the lowest RMSE (3.295 MPa) and highest R2 (0.974). For Flexural and Tensile Strength, it attained R2 values of 0.986 and 0.974, respectively, with 90% and 100% of predictions showing errors below 10%. Among hybrid models, RRSD excelled, with R2 values of 0.94 for Compressive Strength and 0.98 for Flexural and Tensile Strength. These results demonstrate the reliability of machine learning for modeling SFRC properties.

关键词： Experimental concrete samples Steel-fiber-reinforced concrete Strength properties Regression prediction models optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

Asynchronous fully-decentralized SGD in the cluster-based model

引用

THEORETICAL COMPUTER SCIENCE 2025年 1031卷

作者： Attiya, Hagit Schiller, Noa Technion Dept Comp Sci Haifa Israel Tel Aviv Univ Dept Comp Sci Tel Aviv Israel

This paper presents fault-tolerant asynchronous Stochastic Gradient Descent (SGD) algorithms. SGD is widely used for approximating the minimum of a cost function Q , a core part of optimization and learning algorithms. Our algorithms are designed for the cluster-based model, which combines message-passing and shared-memory communication layers. Processes may fail by crashing, and the algorithm inside each cluster is wait-free, using only reads and writes. For a strongly convex Q , our algorithm can withstand partitions of the system. It provides convergence rate that is the maximal distributed acceleration over the optimal convergence rate of sequential SGD. For arbitrary smooth functions, the convergence rate has an additional term that depends on the maximal difference between the parameters at the same iteration. (This holds under standard assumptions on Q .) In this case, the algorithm obtains the same convergence rate as sequential SGD, up to a logarithmic factor. This is achieved by using, at each iteration, a multidimensional approximate agreement algorithm, tailored for the cluster-based model. The general algorithm communicates with nonfaulty processes belonging to clusters that include a majority of all processes. We prove that this condition is necessary when optimizing some non-convex functions.

关键词： optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

A novel high-dimensional sensor calibration framework integrating thermodynamic laws in complex HVAC systems

引用

ENERGY AND BUILDINGS 2025年 327卷

作者： Yan, Chengchu Hu, Kai Xu, Chao Zhuang, Chaoqun Fang, Junjian Gong, Yanfeng Nanjing Tech Univ Coll Urban Construct 200 North Zhongshan Rd Nanjing 210009 Peoples R China China Construct Yipin Investment & Dev Co Ltd Wuhan 430070 Peoples R China

Accurate calibration of sensors is critical for ensuring energy efficient operation of Heating, Ventilation, and Air Conditioning (HVAC) systems in buildings. Due to the high dimensionality of sensor data and the complexity of multiple-fault scenarios, calibrating sensors in large and complex HVAC systems presents significant challenges. To address this issue, this study introduces a novel sensor calibration framework that integrates thermodynamic laws for high-dimensional sensor calibration in complex HVAC systems. The traditional calibration method heavily relies on accurate data, making it difficult to apply in practical engineering projects. The innovative aspect of our method lies in its integration of thermodynamic laws, such as mass balance and energy conservation, with sensor calibration framework. This approach enables the framework to handle high-dimensional sensor measurements effectively without any training data. We compared five optimization algorithms and applied them to a central cooling system in Hong Kong. The results demonstrated that the simulated annealing (SA) is the most robust for solving the calibration problem, even in scenarios with up to 21 faulty sensors, with the calibrated sensor accuracy meeting the standards for conventional chiller plant operations. This novel framework provides a robust and reliable solution for high-dimensional sensor calibration in large and complex HVAC systems, addressing the growing need for precise sensor calibration as the number of installed sensors increases.

关键词： HVAC systems Thermodynamic laws Sensor calibration optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

Comparison of Calibration Strategies for Daily Streamflow Simulations in Semi-Arid Basins

引用

WATER RESOURCES MANAGEMENT 2025年第3期39卷 1089-1105页

作者： Naceur, Khaoula Ait El Khalki, El Mahdi Hadri, Abdessamad Jaffar, Oumar Brocca, Luca Saidi, Mohamed Elmehdi Tramblay, Yves Chehbouni, Abdelghani Mohammed VI Polytech Univ UM6P Int Water Res Inst Benguerir Morocco CNR Res Inst Geohydrol Protect Perugia Italy Cadi Ayyad Univ Fac Sci & Tech Earth Sci Dept Lab L3G Marrakech Morocco Univ Montpellier Espace Dev IRD Montpellier France Mohammed VI Polytech Univ UM6P Ctr Remote Sensing Applicat CRSA BenGuerir Morocco

Hydrological modeling is a crucial tool for water resources management. It becomes more important in data-scarce regions like Morocco. Therefore, accurate parameter tuning of models used in this region is vital for reliable predictions. Traditionally, the Nelder-Mead Simplex Algorithm has been used to calibrate the GR4J and MISDc models. However, this study aims to enhance the calibration process by employing Particle Swarm optimization (PSO), Nelder-Mead Simplex Algorithm (FMIN), Simulated Annealing (SA), and Genetic Algorithm (GA) for daily streamflow forecasts across 26 basins. A sensitivity analysis of their parameters was performed, along with the use of various calibration scenarios. In addition, a snow module was used in the mountainous basins. The research reveals significant sensitivity of the GR4J groundwater exchange coefficient and MISDc parameters related to soil, drainage, and snow dynamics. FMIN and PSO proved to be the most efficient, and the MISDc model outperformed GR4J. The choice of splitting scenario proved critical, and the lower model performance was attributed to discrepancies between calibration and validation periods in terms of runoff coefficients, precipitation-runoff correlations, and the distribution of dry and wet years. Integrating a snow module in both models enhanced their performance in larger basins.

关键词： Hydrological Modeling Morocco GR4J MISDc optimization algorithms Parameter Sensitivity

来源：评论

学校读者我要写书评

暂无评论

A Fixed-Point Iteration Scheme for Sensitivity-Based Distributed Optimal Control

引用

IEEE TRANSACTIONS ON AUTOMATIC CONTROL 2025年第4期70卷 2778-2785页

作者： von Esch, Maximilian Pierer Volz, Andreas Graichen, Knut Friedrich Alexander Univ Erlangen Nurnberg FAU Chair Automatic Control D-91058 Erlangen Germany

This article presents a sensitivity-based algorithm for distributed optimal control problems (OCP) of multi-agent systems with nonlinear dynamics and state/input couplings, as they arise, for instance, in distributed model predictive control. The algorithm relies on first-order sensitivities to cooperatively solve the distributed OCP in parallel. The solutions to the resulting local OCPs are computed with a fixed-point scheme and communicated within one communication step per algorithm iteration to the neighbors. Convergence results are presented under the inexact minimization of the local OCP. The algorithm is evaluated in numerical simulations for an example system.

关键词： Heuristic algorithms Convergence Couplings Costs Trajectory Sensitivity Prediction algorithms Optimal control Vectors Predictive control Agents and autonomous systems distributed optimal control nonlinear systems optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

Leveraging evolutionary algorithms with a dynamic weighted search space approach for fraud detection in healthcare insurance claims

引用

KNOWLEDGE-BASED SYSTEMS 2025年 317卷

作者： Tubishat, Mohammad Tbaishat, Dina Al-Zoubi, Ala'M. Hraiz, Abed-Elalim Habib, Maria Zayed Univ Coll Technol Innovat Abu Dhabi U Arab Emirates Zayed Univ Coll Technol Innovat Dubai U Arab Emirates Al Zaytoonah Univ Jordan Fac Sci & Informat Technol Dept Data Sci & Artificial Intelligence Amman Jordan Univ Jordan King Abdullah II School Informat Technol Amman Jordan Univ Granada Doctoral Sch Sci Technol & Engn Granada Spain

The healthcare industry has been suffering from fraud in many facets for decades, resulting in millions of dollars lost to fictitious claims at the expense of other patients who cannot afford appropriate care. As such, accurately identifying fraudulent claims is one of the most important factors in a well-functioning healthcare system. However, over time, fraud has become harder to detect because of increasingly complex and sophisticated fraud scheme development, data unpreparedness, as well as data privacy concerns. Moreover, traditional methods are proving increasingly inadequate in addressing this issue. To solve this issue a novel evolutionary dynamic weighted search space approach (DW-WOA-SVM) is presented in the current study. The approach has different levels that work simultaneously, where the optimization algorithm is responsible for tuning the Support Vector Machine (SVM) parameters, applying the weighting procedure for the features, and using a dynamic search space to adjust the range values. Tuning the parameters benefits the performance of SVM, and the weighting technique makes it updated with importance and lets the algorithm focus on data structure in addition to optimization objectives. The dynamic search space enhances the search range during the process. Furthermore, large language models have been applied to generate the dataset to improve the quality of the data and address the lack of good dimensionality, helping to enhance the richness of the data. The experiments highlighted the superior performance of this proposed approach than other algorithms.

关键词： Medical fraud Detection Claims SVM LLM Dynamic search space Feature weighing optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

Emerging maximum power point control algorithms for PV system: review, challenges and future trends

引用

ELECTRICAL ENGINEERING 2025年 1-33页

作者： Siddique, Muhammad Abu Bakar Zhao, Dongya Rehman, Ateeq Ur China Univ Petr East China Coll New Energy Qingdao 266580 Peoples R China Saveetha Inst Med & Tech Sci Saveetha Sch Engn Comp Sci & Engn Chennai Tamilnadu India Appl Sci Private Univ Appl Sci Res Ctr Amman Jordan Chandigarh Univ Univ Ctr Res & Dev Mohali 140413 Punjab India

The maximum power point tracking (MPPT) algorithms are essential for ensuring optimal energy conversion and efficient power transfer between the photovoltaic (PV) system and the load. This paper provides a comprehensive review of emerging MPPT algorithms for PV systems under different weather conditions, with a focus on their challenges and future trends. The review covers various types of converters, inverters, MPPT techniques including traditional, optimization, and artificial intelligence (AI)-based control strategies used in PV systems. The converters play a crucial role in converting the DC power generated by the PV panels into usable power that can be consumed by different loads. The paper highlights the working principle, limitations, challenges, and comparison of these techniques to choose most suitable algorithm for a specific application. Furthermore, the review discusses the future trends and enhancement in MPPT algorithms, such as the use of AI and optimization techniques to improve the performance and efficiency of PV systems. Overall, this paper provides valuable perspectives into the current state of MPPT algorithms for PV systems and the potential avenues for future research in this field.

关键词： Maximum power point tracking (MPPT) Artificial intelligence optimization algorithms Renewable energy Power converters Power inverters Energy conversion

来源：评论

学校读者我要写书评

暂无评论

A Hybrid Machine Learning Framework for Early Fault Detection in Power Transformers Using PSO and DMO algorithms

引用

ENERGIES 2025年第8期18卷 2024-2024页

作者： Alenezi, Mohammed Anayi, Fatih Packianather, Michael Shouran, Mokhtar Cardiff Univ Wolfson Ctr Magnet Sch Engn Cardiff CF24 3AA Wales Cardiff Univ Sch Engn High Value Mfg Grp Cardiff CF24 3AA Wales Libyan Ctr Engn Res & Informat Technol Bani Walid Libya

The early detection of faults in power transformers is crucial for ensuring operational reliability and minimizing system disruptions. This study introduces a novel machine learning framework that integrates Particle Swarm optimization (PSO) and Dwarf Mongoose optimization (DMO) algorithms for feature selection and hyperparameter tuning, combined with advanced classifiers such as Decision Trees (DT), Random Forests (RF), and Support Vector Machines (SVM). A 5-fold cross-validation approach was employed to ensure a robust performance evaluation. Feature extraction was performed using both Discrete Wavelet Decomposition (DWD) and Matching Pursuit (MP), providing a comprehensive representation of the dataset comprising 2400 samples and 41 extracted features. Experimental validation demonstrated the efficacy of the proposed framework. The PSO-optimized RF model achieved the highest accuracy of 97.71%, with a precision of 98.02% and an F1 score of 98.63%, followed by the PSO-DT model with a 95.00% accuracy. Similarly, the DMO-optimized RF model recorded an accuracy of 98.33%, with a precision of 98.80% and an F1 score of 99.04%, outperforming other DMO-based classifiers. This novel framework demonstrates significant advancements in transformer protection by enabling accurate and early fault detection, thereby enhancing the reliability and safety of power systems.

关键词： power transformer protection fault detection machine learning classifiers early fault diagnosis optimization algorithms feature extraction

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：