检索结果-内蒙古大学图书馆

Light-EvoOPT: A Lightweight Evolutionary Optimization Framework for Ultra-Large-Scale Mixed Integer Linear Programs

IEEE Transactions on Evolutionary Computation 2025年

作者： Ye, Huigen Xu, Hua Coello, Carlos A. Coello Tsinghua University State Key Laboratory of Intelligent Technology and Systems Department of Computer Science and Technology Beijing100084 China Department of Computer Science D.F México07300 Mexico Tecnologico de Monterrey Faculty of Excellence The School of Engineering and Sciences N.L Monterrey Mexico

Machine Learning (ML)-based optimization frameworks emerge as a promising technique for solving large-scale Mixed Integer Linear Programs (MILPs), as they can capture the mapping between problem structures and optimal solutions to expedite their solution process. However, existing solution frameworks often suffer from high model computation costs, incomplete problem reduction, and reliance on large-scale solvers, leading to performance bottlenecks in ultra-large-scale problems with complex constraints. To address these issues, this paper proposes Light-EvoOPT, a Lightweight Evolutionary Optimization Framework for Ultra-Large-Scale Mixed Integer Linear Programs, which can be divided into four stages: (1) Problem Formulation for problem division to reduce model computational costs, (2) Model-based Initial Solution Prediction for predicting and constructing the initial solution using a small-scale training dataset, (3) Problem Reduction for both variable and constraint reduction, and (4) Evolutionary Optimization for current solution improvement employing a lightweight optimizer. Experiments on four benchmark datasets with tens of millions of variables and constraints and a real-world problem show that the proposed framework based on the sole use of a lightweight optimizer, trained on only one-thousandth of the scale of ultra-large-scale problems, is able to outperform state-of-the-art ML-based frameworks and advanced solvers (e.g. Gurobi) within a specified computational time, validating the feasibility and effectiveness of our proposed ML-based evolutionary optimization framework for ultra-large-scale MILPs. © 2025 IEEE. All rights reserved.

关键词： Mixed-integer linear programming

来源：评论

学校读者我要写书评

暂无评论

DMR: Decomposed Multi-Modality Representations for Frames and Events Fusion in Visual Reinforcement Learning

DMR: Decomposed Multi-Modality Representations for Frames an...

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Haoran Xu Peixi Peng Guang Tan Yuan Li Xinhai Xu Yonghong Tian School of Intelligent Systems Engineering Shenzhen Campus of Sun Yat-sen University Peng Cheng Laboratory School of Electronic and Computer Engineering Shenzhen Graduate School Peking University Academy of Military Sciences School of Computer Science Peking University

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

We explore visual reinforcement learning (RL) using two complementary visual modalities: frame-based RGB cam-era and event-based Dynamic Vision Sensor (DVS). Ex-isting multi-modality visual RL methods often encounter challenges in effectively extracting task-relevant information from multiple modalities while suppressing the in-creased noise, only using indirect reward signals instead of pixel-level supervision. To tackle this, we propose a Decomposed Multi-Modality Representation (DMR) framework for visual RL. It explicitly decomposes the inputs into three distinct components: combined task-relevant features (co-features), RGB-specific noise, and DVS-specific noise. The co-features represent the full information from both modalities that is relevant to the RL task; the two noise components, each constrained by a data reconstruction loss to avoid information leak, are contrasted with the co-features to maximize their difference. Extensive experiments demonstrate that, by explicitly separating the different types of information, our approach achieves substan-tially improved policy performance compared to state-of-the-art approaches.

关键词： Visualization computer vision Noise Reinforcement learning Vision sensors Feature extraction Data mining

来源：评论

学校读者我要写书评

暂无评论

Unknown-Aware Bilateral Dependency Optimization for Defending Against Model Inversion Attacks

引用

IEEE Transactions on Pattern Analysis and Machine Intelligence 2025年 PP卷 PP页

作者： Peng, Xiong Liu, Feng Wang, Nannan Lan, Long Liu, Tongliang Cheung, Yiu-Ming Han, Bo Hong Kong Baptist University Department of Computer Science Hong Kong The University of Melbourne School of Computing and Information Systems Australia Xidian University State Key Laboratory of Integrated Services Networks School of Telecommunications Engineering Shaanxi Xi'an710071 China National University of Defense Technology State Key Laboratory of High Performance Computing School of Computer Hunan Changsha410000 China The University of Sydney Sydney AI Centre School of Computer Science Faculty of Engineering Australia

By abusing access to a well-trained classifier, model inversion (MI) attacks pose a significant threat as they can recover the original training data, leading to privacy leakage. Previous studies mitigated MI attacks by imposing regularization to reduce the dependency between input features and outputs during classifier training, a strategy known as unilateral dependency optimization. However, this strategy contradicts the objective of minimizing the supervised classification loss, which inherently seeks to maximize the dependency between input features and outputs. Consequently, there is a trade-off between improving the model's robustness against MI attacks and maintaining its classification performance. To address this issue, we propose the bilateral dependency optimization strategy (BiDO), a dual-objective approach that minimizes the dependency between input features and latent representations, while simultaneously maximizing the dependency between latent representations and labels. BiDO is remarkable for its privacy-preserving capabilities. However, models trained with BiDO exhibit diminished capabilities in out-of-distribution (OOD) detection compared to models trained with standard classification supervision. Given the open-world nature of deep learning systems, this limitation could lead to significant security risks, as encountering OOD inputs—whose label spaces do not overlap with the in-distribution (ID) data used during training-is inevitable. To address this, we leverage readily available auxiliary OOD data to enhance the OOD detection performance of models trained with BiDO. This leads to the introduction of an upgraded framework, unknown-aware BiDO (BiDO+), which mitigates both privacy and security concerns. As a highlight, with comparable model utility, BiDO-HSIC+ reduces the FPR95 by 55.02% and enhances the AUCROC by 9.52% compared to BiDO-HSIC, while also providing superior MI robustness. © 1979-2012 IEEE.

关键词： Differential privacy

来源：评论

学校读者我要写书评

暂无评论

PEIE: Physics Embedded Illumination Estimation for Adaptive Dehazing 39

PEIE: Physics Embedded Illumination Estimation for Adaptive ...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Liu, Huaizhuo Hu, Hai-Miao Jiang, Yonglong Liu, Yurui Hangzhou Innovation Institute Beihang University China State Key Laboratory of Virtual Reality Technology and Systems Beihang University China School of Computer Science and Engineering Beihang University China

ISBN: (纸本)157735897X

Deep learning-based methods have made significant progress in image dehazing. However, these methods often falter when applied to real-world hazy images, primarily due to the scarcity of paired real-world data and the limitations of current dehazing feature extractors. Toward these issues, we introduce a novel Physics Embedded Illumination Estimation (PEIE) method for adaptive real-world dehazing. Specifically, (1) we identify the limitations of the widely used Atmospheric Scattering Model and propose a new physical model, the Illumination-Adaptive Scattering Model (IASM), for more accurate illumination representation in hazy imaging;(2) we develop a robust data synthesis pipeline that leverages the physics embedded illumination estimation to generate realistic hazy images;and (3) we design an Illumination-Adaptive Dehazing Unit (IDU) to extract dehazing features consistent with our proposed IASM in the latent space. By integrating the IDU into a U-Net architecture to create IADNet, we achieve significant improvements in dehazing performance through end-to-end training on synthetic data. Extensive experiments validate the superior performance of our PEIE method, significantly surpassing the state-of-the-arts in real-world dehazing. © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

BMP-SD: Marrying Binary and Mixed-Precision Quantization for Efficient Stable Diffusion Inference

BMP-SD: Marrying Binary and Mixed-Precision Quantization for...

引用

Design, Automation and Test in Europe Conference and Exhibition

作者： Cheng Gu Gang Li Xiaolong Lin Jiayao Ling Jian Cheng Xiaoyao Liang Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai China The Key Laboratory of Cognition and Decision Intelligence for Complex Systems CASIA Beijing China AiRiA Nanjing China

ISBN: (数字)9783982674100

ISBN: (纸本)9798331534646

Stable Diffusion (SD) is an emerging deep neural network (DNN) model that has demonstrated impressive capabilities in generative tasks such as text-to-image generation. However, the iterative denoising stage of the SD model is extremely expensive in both computations and memory accesses, making it challenging for fast and energy-efficient edge deployment. To alleviate the overhead of denoising, in this paper we propose BMP-SD, a post-training quantization framework for hardware-efficient SD inference. BMP-SD employs binary weight quantization to significantly reduce the computational complexity and memory footprint of iterative denoising, along with dynamic step-aware mixed-precision activation quantization, based on the observation that not all denoising steps are equally important for a specific input prompt. Experiments on the text-to-image generation task show that BMP-SD achieves mixed-precision (W1.73A4.87) with minimal accuracy loss on MS-COCO 2014 dataset. We also evaluate the BMP-SD quantized model on three state-of-the-art bit-flexible DNN accelerators, results reveal that our method can deliver up to 5.14× performance and 3.85×energy efficiency improvements compared to W8A8 quantization.

关键词： Quantization (signal) Accuracy Computational modeling Noise reduction Text to image Europe Artificial neural networks Energy efficiency Iterative methods Computational complexity

来源：评论

学校读者我要写书评

暂无评论

SBQ: Exploiting Significant Bits for Efficient and Accurate Post-Training DNN Quantization

SBQ: Exploiting Significant Bits for Efficient and Accurate ...

引用

2025 Design, Automation and Test in Europe Conference, DATE 2025

作者： Ling, Jiayao Li, Gang Hu, Qinghao Lin, Xiaolong Gu, Cheng Cheng, Jian Liang, Xiaoyao Shanghai Jiao Tong University Department of Computer Science and Engineering Shanghai China Casia The Key Laboratory of Cognition and Decision Intelligence for Complex Systems Beijing China AiRiA Nanjing China

ISBN: (纸本)9783982674100

Post-Training Quantization is an effective technique for deep neural network acceleration. However, as the bit-width decreases to 4 bits and below, PTQ faces significant challenges in preserving accuracy, especially for attention-based models like LLMs. The main issue lies in considerable clipping and rounding errors induced by the limited number of quantization levels and narrow data range in conventional low-precision quantization. In this paper, we present an efficient and accurate PTQ method that targets 4 bits and below through algorithm and architecture co-design. Our key idea is to dynamically extract a small portion of significant bit terms from high-precision operands to perform low-precision multiplications under the given computational budget. Specifically, we propose Significant-Bit Quantization (SBQ). It exploits a product-aware method to dynamically identify significant terms and an error-compensated computation scheme to minimize compute errors. We present a dedicated inference engine to unleash the power of SBQ. Experiments on CNNs, ViTs, and LLMs reveal that SBQ consistently outperforms prior PTQ methods under 24-bit quantization. We also compare the proposed inference engine with state-of-the-art bit-operation-based quantization architectures TQ and Sibia. Results show that SBQ can achieve the highest area and energy efficiency. © 2025 EDAA.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

SBQ: Exploiting Significant Bits for Efficient and Accurate Post-Training DNN Quantization

SBQ: Exploiting Significant Bits for Efficient and Accurate ...

引用

Design, Automation and Test in Europe Conference and Exhibition

作者： Jiayao Ling Gang Li Qinghao Hu Xiaolong Lin Cheng Gu Jian Cheng Xiaoyao Liang Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai China The Key Laboratory of Cognition and Decision Intelligence for Complex Systems CASIA Beijing China AiRiA Nanjing China

ISBN: (数字)9783982674100

ISBN: (纸本)9798331534646

Post-Training Quantization is an effective technique for deep neural network acceleration. However, as the bit-width decreases to 4 bits and below, PTQ faces significant challenges in preserving accuracy, especially for attention-based models like LLMs. The main issue lies in considerable clipping and rounding errors induced by the limited number of quantization levels and narrow data range in conventional low-precision quantization. In this paper, we present an efficient and accurate PTQ method that targets 4 bits and below through algorithm and architecture co-design. Our key idea is to dynamically extract a small portion of significant bit terms from high-precision operands to perform low-precision multiplications under the given computational budget. Specifically, we propose Significant-Bit Quantization (SBQ). It exploits a product-aware method to dynamically identify significant terms and an error-compensated computation scheme to minimize compute errors. We present a dedicated inference engine to unleash the power of SBQ. Experiments on CNNs, ViTs, and LLMs reveal that SBQ consistently outperforms prior PTQ methods under 2~4-bit quantization. We also compare the proposed inference engine with state-of-the-art bit-operation-based quantization architectures TQ and Sibia. Results show that SBQ can achieve the highest area and energy efficiency.

关键词： Quantization (signal) Accuracy Heuristic algorithms Europe computer architecture Artificial neural networks Energy efficiency Inference algorithms Engines Faces

来源：评论

学校读者我要写书评

暂无评论

Noninvasive Tracking of Every Individual in Unmarked Mouse Groups Using Multi-Camera Fusion and Deep Learning

引用

Neuroscience Bulletin 2023年第6期39卷 893-910页

作者： Feng Su Yangzhen Wang Mengping Wei Chong Wang Shaoli Wang Lei Yang Jianmin Li Peijiang Yuan Dong-Gen Luo Chen Zhang Department of Neurobiology School of Basic Medical SciencesBeijing Key Laboratory of Neural Regeneration and RepairCapital Medical UniversityBeijing100069China Chinese Institute for Brain Research Beijing102206China State Key Laboratory of Translational Medicine and Innovative Drug Development Nanjing210000China Peking-Tsinghua Center for Life Sciences Academy for Advanced Interdisciplinary StudiesPeking UniversityBeijing100871China School of Life Sciences Tsinghua UniversityBeijing100084China School of Biological Science and Medical Engineering Beihang UniversityBeijing100191China The Key Laboratory of Developmental Genes and Human Disease Institute of Life SciencesSoutheast UniversityNanjing210096JiangsuChina Institute for Artificial Intelligence the State Key Laboratory of Intelligence Technology and SystemsBeijing National Research Center for Information Science and TechnologyDepartment of Computer Science and TechnologyTsinghua UniversityBeijing100084China School of Mechanical Engineering and Automation Beihang UniversityBeijing100191China

Accurate and efficient methods for identifying and tracking each animal in a group are needed to study complex behaviors and social *** tracking methods(e.g.,marking each animal with dye or surgically implanting microchips)can be invasive and may have an impact on the social behavior being *** overcome these shortcomings,video-based methods for tracking unmarked animals,such as fruit flies and zebrafish,have been ***,tracking individual mice in a group remains a challenging problem because of their flexible body and complicated interaction *** this study,we report the development of a multi-object tracker for mice that uses the Faster region-based convolutional neural network(R-CNN)deep learning algorithm with geometric transformations in combination with multi-camera/multi-image fusion *** system successfully tracked every individual in groups of unmarked mice and was applied to investigate chasing *** proposed system constitutes a step forward in the noninvasive tracking of individual mice engaged in social behavior.

关键词： Noninvasive tracking Deep learning Multi-camera Mouse group Social interaction

来源：评论

学校读者我要写书评

暂无评论

Cyber Security Intrusion Detection for Agriculture 4.0: Machine Learning-Based Solutions, Datasets,and Future Directions

引用

IEEE/CAA Journal of Automatica Sinica 2022年第3期9卷 407-436页

作者： Mohamed Amine Ferrag Lei Shu Othmane Friha Xing Yang Department of Computer Science Guelma UniversityB.P.40124000Algeria IEEE College of Artificial Intelligence Nanjing Agricultural UniversityNanjing 210031China School of Engineering University of LincolnLincoln LN67TSUK Networks and Systems Laboratory(LRS) University of Badji Mokhtar-AnnabaB.P.12Annaba 23000Algeria College of Engineering Nanjing Agricultural UniversityNanjing 210031China

In this paper,we review and analyze intrusion detection systems for Agriculture 4.0 cyber ***,we present cyber security threats and evaluation metrics used in the performance evaluation of an intrusion detection system for Agriculture ***,we evaluate intrusion detection systems according to emerging technologies,including,Cloud computing,Fog/Edge computing,Network virtualization,Autonomous tractors,Drones,Internet of Things,Industrial agriculture,and Smart *** on the machine learning technique used,we provide a comprehensive classification of intrusion detection systems in each emerging ***,we present public datasets,and the implementation frameworks applied in the performance evaluation of intrusion detection systems for Agriculture ***,we outline challenges and future research directions in cyber security intrusion detection for Agriculture 4.0.

关键词： Agriculture 4.0 cyber security intrusion detectionsystem machine learning approaches smart agriculture

来源：评论

学校读者我要写书评

暂无评论

OUTLIER SYNTHESIS VIA HAMILTONIAN MONTE CARLO FOR OUT-OF-DISTRIBUTION DETECTION

arXiv

引用

arXiv 2025年

作者： Li, Hengzhuang Zhang, Teng National Engineering Research Center for Big Data Technology and System Service Computing Technology and Systems Laboratory Cluster and Grid Computing Lab School of Computer Science and Technology Huazhong University of Science and Technology Wuhan China

Out-of-distribution (OOD) detection is crucial for developing trustworthy and reliable machine learning systems. Recent advances in training with auxiliary OOD data demonstrate efficacy in enhancing detection capabilities. Nonetheless, these methods heavily rely on acquiring a large pool of high-quality natural outliers. Some prior methods try to alleviate this problem by synthesizing virtual outliers but suffer from either poor quality or high cost due to the monotonous sampling strategy and the heavy-parameterized generative models. In this paper, we overcome all these problems by proposing the Hamiltonian Monte Carlo Outlier Synthesis (HamOS) framework, which views the synthesis process as sampling from Markov chains. Based solely on the in-distribution data, the Markov chains can extensively traverse the feature space and generate diverse and representative outliers, hence exposing the model to miscellaneous potential OOD scenarios. The Hamiltonian Monte Carlo with sampling acceptance rate almost close to 1 also makes our framework enjoy great efficiency. By empirically competing with SOTA baselines on both standard and large-scale benchmarks, we verify the efficacy and efficiency of our proposed HamOS. Our code is available at: https://***/Fir-lat/HamOS_OOD. © 2025, CC BY.

关键词： Markov chains

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

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

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

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

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：