检索结果-内蒙古大学图书馆

Reliability Assessment of Large DNN Models: Trading Off Performance and Accuracy

学校读者我要写书评

暂无评论

Reliability Assessment of Large DNN Models: Trading Off Perf...

IFIP International Conference on Very Large Scale Integration (VLSI-SoC)

作者： Junchao Chen Giuseppe Esposito Fernando Fernandes dos Santos Juan-David Guerrero-Balaguera Angeliki Kritikakou Milos Krstic Robert Limas Sierra Josie E. Rodriguez Condia Matteo Sonza Reorda Marcello Traiola Alessandro Veronesi IHP - Leibniz Institute for High-Performance Microelectronics Germany Department of Control and Computer Engineering (DAUIN) Politecnico di Torino Turin Italy Univ Rennes CNRS Inria IRISA - UMR 6074 Rennes France University of Potsdam Germany

ISBN: (数字)9798331539672

ISBN: (纸本)9798331539689

The adoption of Deep Neural Networks (DNNs) in several domains allows for increased effectiveness in applications that deal with massive data-intensive and complex data inputs. When employed in safety-critical scenarios, such as automotive, aerospace, healthcare, and autonomous robotics, assessing the DNNs' reliability and functional safety is crucial to ensure their correct in-field operation, even in the presence of hardware faults. However, the system complexity and the massive amounts of data to be processed by DNNs prevent the effective adoption of traditional strategies for reliability characterization and for identifying the most fault-sensitive structures. Accurate fault assessment strategies usually require unacceptable computational power and large evaluation times. On the other hand, faster strategies commonly lack accuracy in correctly representing system faults. Consequently, it is necessary to develop effective strategies that trade-off between performance and accuracy. This work analyses three reliability assessment strategies for deep neural networks and their underlying hardware, highlighting the main solutions and challenges in terms of evaluation performance and fault characterization accuracy. We overview different solutions to evaluate the hardware accelerators implementing DNNs at three abstraction levels: $i$ ) by physically injecting faults on a GPU running DNNs, ii) by performing microarchitectural characterization of GPUs to develop application-accurate error models, and iii) by using structure-aware cross-layer error modeling on DNN hardware accelerators. Our experimental results indicate that accurate error representation requires structural features from the targeted hardware.

关键词：

Multi-task multi-constraint differential evolution with elite-guided knowledge transfer for coal mine integrated energy system dispatching

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Dai, Canyun Sun, Xiaoyan Hu, Hejuan Song, Wei Zhang, Yong Gong, Dunwei School of Information and Control Engineering China University of Mining and Technology Xuzhou China School of Artificial Intelligence and Computer Science Jiangnan University Wuxi China School of Information Science and Technology Qingdao University of Science and Technology Qingdao China

The dispatch optimization of coal mine integrated energy system is challenging due to high dimensionality, strong coupling constraints, and multi-objective. Existing constrained multi-objective evolutionary algorithms struggle with locating multiple small and irregular feasible regions, making them inapplicable to this problem. To address this issue, we here develop a multi-task evolutionary algorithm framework that incorporates the dispatch-correlated domain knowledge to effectively deal with strong constraints and multi-objective optimization. Possible evolutionary multi-task construction strategy based on complex constraint relationship analysis and handling, i.e., constraint-coupled spatial decomposition, constraint strength classification and constraint handling technique, is first explored. Within the multi-task evolutionary optimization framework, two strategies, i.e., an elite-guided knowledge transfer by designing a special crowding distance mechanism to select dominant individuals from each task, and an adaptive neighborhood technology-based mutation to effectively balance the diversity and convergence of each optimized task for the differential evolution algorithm, are further developed. The performance of the proposed algorithm in feasibility, convergence, and diversity is demonstrated in a case study of a coal mine integrated energy system by comparing with CPLEX solver and seven state-of-the-art constrained multiobjective evolutionary algorithms. Copyright © 2024, The Authors. All rights reserved.

关键词： Coal mines

Iteration Complexity of a Fixed-Stepsize SQP Method for Nonconvex Optimization with Convex Constraints 5th

学校读者我要写书评

暂无评论

Iteration Complexity of a Fixed-Stepsize SQP Method for Nonc...

5th International Conference on Numerical Analysis and Optimization: Theory, Methods, Applications and Technology Transfer, NAOV 2020

作者： Facchinei, Francisco Kungurtsev, Vyacheslav Lampariello, Lorenzo Scutari, Gesualdo Department of Computer Control and Management Engineering Antonio Ruberti Sapienza University of Rome Rome Italy Department of Computer Science Faculty of Electrical Engineering Czech Technical University in Prague Prague Czech Republic Department of Business Studies Roma Tre University Rome Italy School of Industrial Engineering Purdue University West-LafayetteIN United States

ISBN: (纸本)9783030720391

We consider an SQP method for solving nonconvex optimization problems whose feasible set is convex and with an objective function that is the sum of a smooth nonconvex term and a nonsmooth, convex one. In the proposed method, at each iteration, a direction is generated by solving a strongly convex approximation to the original problem and then a fixed-stepsize is taken in that direction. The complexity result we establish is, as far as we are aware, the best available for the rather general setting we consider. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Fixed-stepsize Iteration complexity SQP methods

Machine Learning for Predicted Extration Rate of Volatile Oil from Biomass Through Supercritical Fluid Extration Technology

学校读者我要写书评

暂无评论

SSRN

SSRN 2024年

作者： Shen, Boxiong Xiong, Lifu Tang, Cheng Liu, Shunyao Tianjin Key Laboratory of Clean Energy and Pollution Control School of Energy and Environmental Engineering Hebei University of Technology Tianjin300401 China School of Computer Science and Engineering Hunan University of Information Technology Changsha410151 China China Academic of Electronics and Information Technology Beijing100041 China

This paper utilizes modern artificial intelligence technology to evaluate and predict the practical application of supercritical fluid extraction technology. This method offers an appropriate opportunity to develop sustainable and environmentally friendly technologies, which hold valuable advantages in the industrial extraction sector. We simulated the extraction rate of biomass volatile oil using multiple machine learning models with supercritical carbon dioxide as solvent. The input was the pressure and temperature in the dataset used, and the single output was the extraction rate. Three well-known supervised learning models, Kernel Extreme Learning Machine, Long Short-Term Memory, and Support Vector Machine were selected for this study to model the process. Hyperparameters of these models were optimized using the Whale Optimization Algorithm and then the final models were evaluated using multiple metrics. In terms of the coefficient of determination, the scores of KELM, LSTM, and SVM models were 0.997, 0.982, and 0.963, respectively. In addition, their error rates were 1.75, 8.98, and 10.25 when using the Mean Absolute Error metric. Finally, the KELM model was selected as the best model, and the results of optimizing extraction parameters were discussed. Trend of how temperature and pressure affect the extraction rate was also discussed using the best model. © 2024, The Authors. All rights reserved.

关键词： Supercritical fluid extraction

Thunder: Thumbnail based Fast Lightweight Image Denoising Network

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Zhou, Yifeng Xu, Xing Liu, Shuaicheng Wang, Guoqing Lu, Huimin Shen, Heng Tao Center for Future Multimedia School of Computer Science and Engineering University of Electronic Science and Technology of China China Department of Mechanical and Control Engineering Kyushu Institute of Technology Japan

To achieve promising results on removing noise from real-world images, most of existing denoising networks are formulated with complex network structure, making them impractical for deployment. Some attempts focused on reducing the number of filters and feature channels but suffered from large performance loss, and a more practical and lightweight denoising network with fast inference speed is of high demand. To this end, a Thumbnail based Denoising Network dubbed Thunder, is proposed and implemented as a lightweight structure for fast restoration without comprising the denoising capabilities. Specifically, the Thunder model contains two newly-established modules: (1) a wavelet-based Thumbnail Subspace Encoder (TSE) which can leverage sub-bands correlation to provide an approximate thumbnail based on the low-frequent feature;(2) a Subspace Projection based Refine Module (SPR) which can restore the details for thumbnail progressively based on the subspace projection approach. Extensive experiments have been carried out on two real-world denoising benchmarks, demonstrating that the proposed Thunder outperforms the existing lightweight models and achieves competitive performance on PSNR and SSIM when compared with the complex designs. Copyright © 2022, The Authors. All rights reserved.

关键词： Restoration

Fully passive measurement-device-independent quantum key distribution

学校读者我要写书评

暂无评论

Physical Review Applied 2024年第6期21卷 064056-064056页

作者： Jinjie Li Wenyuan Wang Hoi-Kwong Lo Department of Physics HK Institute of Quantum Science & Technology Dept. of Electrical & Computer Engineering Centre for Quantum Information and Quantum Control (CQIQC) Dept. of Physics Quantum Bridge Technologies Inc. 100 College Street Toronto ON M5G 1L5 Canada

A recently proposed fully passive quantum key distribution (QKD) removes all source modulator side channels. In this work, we combine fully passive sources with measurement-device-independent (MDI) QKD to simultaneously remove side channels from source modulators and detectors. We show a numerical simulation of passive MDI QKD, and we obtain an acceptable key rate while achieving much better implementation security, as well as ease of implementation, compared with a recently proposed fully passive twin-field QKD, paving the way towards more secure and practical QKD systems. We prove that a fully passive protocol is compatible with MDI QKD and we propose a novel idea that can improve the sifting efficiency.

关键词： Optical quantum information processing Photonics Quantum communication, protocols & technology Quantum cryptography

Ultra-compact Binary Neural Networks for Human Activity Recognition on RISC-V Processors

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Daghero, Francesco Xie, Chen Pagliari, Daniele Jahier Burrello, Alessio Castellano, Marco Gandolfi, Luca Calimera, Andrea Macii, Enrico Poncino, Massimo Department of Control and Computer Engineering Politecnico di Torino Turin Italy Department of Electrical Electronic and Information Engineering University of Bologna Bologna Italy STMicroelectronics Cornaredo Italy

Human Activity Recognition (HAR) is a relevant inference task in many mobile applications. State-of-the-art HAR at the edge is typically achieved with lightweight machine learning models such as decision trees and Random Forests (RFs), whereas deep learning is less common due to its high computational complexity. In this work, we propose a novel implementation of HAR based on deep neural networks, and precisely on Binary Neural Networks (BNNs), targeting low-power general purpose processors with a RISC-V instruction set. BNNs yield very small memory footprints and low inference complexity, thanks to the replacement of arithmetic operations with bit-wise ones. However, existing BNN implementations on general purpose processors impose constraints tailored to complex computer vision tasks, which result in over-parametrized models for simpler problems like HAR. Therefore, we also introduce a new BNN inference library, which targets ultra-compact models explicitly. With experiments on a single-core RISC-V processor, we show that BNNs trained on two HAR datasets obtain higher classification accuracy compared to a state-of-the-art baseline based on RFs. Furthermore, our BNN reaches the same accuracy of a RF with either less memory (up to 91%) or more energy-efficiency (up to 70%), depending on the complexity of the features extracted by the RF. © 2022, CC BY.

关键词： Energy efficiency

MD-SLAM: Multi-cue Direct SLAM

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Di Giammarino, Luca Brizi, Leonardo Guadagnino, Tiziano Stachniss, Cyrill Grisetti, Giorgio The Department of Computer Control and Management Engineering Antonio Ruberti Sapienza University of Rome Italy The University of Bonn Germany The Department of Engineering Science The University of Oxford United Kingdom

Simultaneous Localization and Mapping (SLAM) systems are fundamental building blocks for any autonomous robot navigating in unknown environments. The SLAM implementation heavily depends on the sensor modality employed on the mobile platform. For this reason, assumptions on the scene’s structure are often made to maximize estimation accuracy. This paper presents a novel direct 3D SLAM pipeline that works independently for RGB-D and LiDAR sensors. Building upon prior work on multi-cue photometric frame-to-frame alignment [4], our proposed approach provides an easy-to-extend and generic SLAM system. Our pipeline requires only minor adaptations within the projection model to handle different sensor modalities. We couple a position tracking system with an appearance-based relocalization mechanism that handles large loop closures. Loop closures are validated by the same direct registration algorithm used for odometry estimation. We present comparative experiments with state-of-the-art approaches on publicly available benchmarks using RGB-D cameras and 3D LiDARs. Our system performs well in heterogeneous datasets compared to other sensor-specific methods while making no assumptions about the environment. Finally, we release an open-source C++ implementation of our system. Copyright © 2022, The Authors. All rights reserved.

关键词： Pipelines

Exploring the Economic Feasibility of Advanced Air Mobility in the Early Stages

学校读者我要写书评

暂无评论

IEEE Transactions on Intelligent Vehicles

IEEE Transactions on Intelligent Vehicles 2024年第5期9卷 4826-4830页

作者： Jingqiu Guo Long Chen Lingxi Li Xiaoxiang Na Ljubo Vlacic Fei-Yue Wang Key Laboratory of Road and Traffic Engineering of the Ministry of Education Tongji University Shanghai China State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing China Department of Electrical and Computer Engineering Purdue School of Engineering and Technology Indiana University–Purdue University Indianapolis (IUPUI) Indianapolis IN USA Department of Engineering University of Cambridge Cambridge U.K. Institute of Intelligent and Integrated Systems and the School of Engineering and Built Environment Griffith University Nathan QLD Australia State Key Laboratory for Management and Control of Complex Systems Institute of Automation Chinese Academy of Sciences Beijing China Institute of Engineering Macau University of Science and Technology Macau China

Advanced Air Mobility (AAM) envisages a sustainable, safe, convenient, and affordable air transport system. In socio-technical transition of AAM, there are a number of trade-offs in ecosystem that need to be studied. Three perspectives on economic feasibility are explored: first, based on history of VTOL services and value of time estimates, we discuss whether AAM can provide customers with competitive mobility services; second, what are the stakeholders’ insights on the deployment of AAM; last, the experience in the development of autonomous driving technology, such as parallel intelligence, can inform future AAM research.

关键词： Active appearance model Costs Economics Transportation Helicopters Aircraft propulsion Public transportation