检索结果-内蒙古大学图书馆

Estimator Meets Equilibrium Perspective: A Rectified Straight Through Estimator for Binary Neural Networks Training

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Wu, Xiao-Ming Zheng, Dian Liu, Zuhao Zheng, Wei-Shi School of Computer Science and Engineering Sun Yat-sen University China Pengcheng Lab China Guangdong Province Key Laboratory of Information Security Technology China Key Laboratory of Machine Intelligence and Advanced Computing Ministry of Education China

Binarization of neural networks is a dominant paradigm in neural networks compression. The pioneering work BinaryConnect uses Straight Through Estimator (STE) to mimic the gradients of the sign function, but it also causes the crucial inconsistency problem. Most of the previous methods design different estimators instead of STE to mitigate it. However, they ignore the fact that when reducing the estimating error, the gradient stability will decrease concomitantly. These highly divergent gradients will harm the model training and increase the risk of gradient vanishing and gradient exploding. To fully take the gradient stability into consideration, we present a new perspective to the BNNs training, regarding it as the equilibrium between the estimating error and the gradient stability. In this view, we firstly design two indicators to quantitatively demonstrate the equilibrium phenomenon. In addition, in order to balance the estimating error and the gradient stability well, we revise the original straight through estimator and propose a power function based estimator, Rectified Straight Through Estimator (ReSTE for short). Comparing to other estimators, ReSTE is rational and capable of flexibly balancing the estimating error with the gradient stability. Extensive experiments on CIFAR-10 and ImageNet datasets show that ReSTE has excellent performance and surpasses the state-of-the-art methods without any auxiliary modules or losses. Copyright © 2023, The Authors. All rights reserved.

关键词： Stability

Repairing with Zero Skip Cost

学校读者我要写书评

暂无评论

Repairing with Zero Skip Cost

IEEE International Symposium on information Theory

作者： Yeow Meng Chee Son Hoang Dau Tuvi Etzion Han Mao Kiah Yuan Luo Wenqin Zhang Dept. of Industrial Systems Engineering and Management National University of Singapore School of Computing Technologies RMIT University Melbourne VIC Australia Computer Science Faculty Technion Israel Institute of Technology School of Physical and Mathematical Sciences Nanyang Technological University Singapore School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai China

To measure repair latency at helper nodes, we introduce a new metric called skip cost that quantifies the number of contiguous sections accessed on a disk. We provide explicit constructions of zigzag codes and fractio... 详细信息

ISBN: (数字)9798350382846

ISBN: (纸本)9798350382853

关键词： Measurement Costs Codes Maintenance engineering

A Framework for the Adoption of Agile within Software SMEs in Saudi Arabia 21

学校读者我要写书评

暂无评论

A Framework for the Adoption of Agile within Software SMEs i...

Proceedings of the 2021 European Symposium on Software Engineering

作者： Fahad Altuwaijri Maria Angela Ferrario School of Computing and Communications Lancaster University UK and Department of Information TechnologyCollege of Computer Qassim University Saudi Arabia School of Computing and Communications Lancaster University UK and School of ElectronicsElectrical Engineering and Computer Science Queen's University UK

ISBN: (纸本)9781450385060

Agile is an established and globally adopted software development methodology for managing software projects, known for reducing time to market and costs, and improving customer satisfaction. However, in developing countries, such as those in the Middle East, Agile is still in the early stages of adoption compared to Europe and North America. This is problematic, specifically in the case of Saudi Arabia, which is increasingly looking at the software industry to diversify its economy to reduce the reliance on the oil industry. It is essential to understand the factors that influence the decision of Saudi software organisations to adopt Agile. Hence, this study aims to critically review the influential factors identified in the Agile adoption framework that may support or challenge the adoption of Agile in small and medium-sized enterprises (SMEs) in Saudi Arabia. To this end, we carried out a focus group study with five software practitioners working in different Saudi software SMEs. The main findings indicate the apparent influence of all the identified factors in the developed framework except those related to process, with a high level of importance attached to culture, people and knowledge factors. This study refines the framework and helps software organisations assess their readiness for adopting Agile during the development process.

关键词： software engineering Agile software development Agile adoption framework Adoption SMEs Agile methods Agile

SegAN for recognition of caries from 2D-panoramic X-ray images

学校读者我要写书评

暂无评论

IEEE Access 2025年 13卷 100419-100432页

作者： Srinivasu, Parvathanenu Naga Aruna Kumari, Gorli L. Jaya Kumari, D. Barsocchi, Paolo Bhoi, Akash Kumar Amrita Vishwa Vidyapeetham Amrita School of Computing Andhra Pradesh Amaravati 522503 India GITAM University Department of Computer Science and Engineering GST Andhra Pradesh Visakhapatnam 530045 India Department of Computer Science and Engineering Sri Vasavi Engineering College Andhra Pradesh Tadepalligudem India National Research Council Institute of Information Science and Technologies Pisa 56124 Italy Sikkim Manipal University Sikkim Manipal Institute of Technology Sikkim Gangtok 737102 India

Accurate recognition and segmentation of dental caries in 2D panoramic X-ray images are crucial for timely diagnosis and strategic treatment planning. The current study uses a Generative Adversarial Network (GAN) model named SegAN to segment the X-ray samples to recognize the caries. The SegAN model works in an adversarial architecture in which the generator focuses on creating precise segmentation maps of caries from 2D panoramic X-ray images. On the other hand, the discriminator ensures that the output matches realistic segmentation patterns. The SegAN model efficiently handles the local and global contextual information for precise segmentation by considering Pixel-Wise and structural loss measures that assist in better segmentation of complex structures. Moreover, the SegAN model efficiently deals with noisy data and effectively handles class imbalances. Data augmentation, like histogram equalization and affine transforms, is performed on the input images for precise segmentation of the samples. The model was evaluated on both raw and preprocessed dental X-ray images using standard quantitative metrics. SegAN demonstrated superior performance compared to traditional segmentation approaches, achieving an accuracy of 98.5% and a dice coefficient of 0.936 in caries detection. © 2013 IEEE.

关键词： Data Augmentation Deep Learning Dental imaging Image Segmentation X-ray Images

ASAG: Building Strong One-Decoder-Layer Sparse Detectors via Adaptive Sparse Anchor Generation

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Fu, Shenghao Yan, Junkai Gao, Yipeng Xie, Xiaohua Zheng, Wei-Shi School of Computer Science and Engineering Sun Yat-sen University China Pengcheng Lab China Guangdong Province Key Laboratory of Information Security Technology China Key Laboratory of Machine Intelligence and Advanced Computing Ministry of Education China

Recent sparse detectors with multiple, e.g. six, decoder layers achieve promising performance but much inference time due to complex heads. Previous works have explored using dense priors as initialization and built one-decoder-layer detectors. Although they gain remarkable acceleration, their performance still lags behind their six-decoder-layer counterparts by a large margin. In this work, we aim to bridge this performance gap while retaining fast speed. We find that the architecture discrepancy between dense and sparse detectors leads to feature conflict, hampering the performance of one-decoder-layer detectors. Thus we propose Adaptive Sparse Anchor Generator (ASAG) which predicts dynamic anchors on patches rather than grids in a sparse way so that it alleviates the feature conflict problem. For each image, ASAG dynamically selects which feature maps and which locations to predict, forming a fully adaptive way to generate image-specific anchors. Further, a simple and effective Query Weighting method eases the training instability from adaptiveness. Extensive experiments show that our method outperforms dense-initialized ones and achieves a better speed-accuracy tradeoff. The code is available at https://***/iSEE-Laboratory/ASAG. Copyright © 2023, The Authors. All rights reserved.

关键词： Decoding

ASAG: Building Strong One-Decoder-Layer Sparse Detectors via Adaptive Sparse Anchor Generation

学校读者我要写书评

暂无评论

ASAG: Building Strong One-Decoder-Layer Sparse Detectors via...

International Conference on Computer Vision (ICCV)

作者： Shenghao Fu Junkai Yan Yipeng Gao Xiaohua Xie Wei-Shi Zheng School of Computer Science and Engineering Sun Yat-sen University China Guangdong Province Key Laboratory of Information Security Technology China Key Laboratory of Machine Intelligence and Advanced Computing Ministry of Education China Pengcheng Lab China

Recent sparse detectors with multiple, e.g. six, decoder layers achieve promising performance but much inference time due to complex heads. Previous works have explored using dense priors as initialization and built one-decoder-layer detectors. Although they gain remarkable acceleration, their performance still lags behind their six-decoder-layer counterparts by a large margin. In this work, we aim to bridge this performance gap while retaining fast speed. We find that the architecture discrepancy between dense and sparse detectors leads to feature conflict, hampering the performance of one-decoder-layer detectors. Thus we propose Adaptive Sparse Anchor Generator (ASAG) which predicts dynamic anchors on patches rather than grids in a sparse way so that it alleviates the feature conflict problem. For each image, ASAG dynamically selects which feature maps and which locations to predict, forming a fully adaptive way to generate image-specific anchors. Further, a simple and effective Query Weighting method eases the training instability from adaptiveness. Extensive experiments show that our method outperforms dense-initialized ones and achieves a better speed-accuracy trade-off. The code is available at https://***/iSEE-Laboratory/ASAG.

关键词：

BackCache: Mitigating Contention-Based Cache Timing Attacks by Hiding Cache Line Evictions

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Wang, Quancheng Zhang, Xige Wang, Han Gu, Yuzhe Tang, Ming Key Laboratory of Aerospace Information Security and Trusted Computing Ministry of Education School of Cyber Science and Engineering Wuhan University Wuhan430072 China Beijing Smartchip Microelectronics Technology Co. Ltd. Beijing100192 China

Caches are used to reduce the speed gap between the CPU and memory to improve the performance and efficiency of modern processors. However, attackers can use contention-based cache timing attacks to steal sensitive information from victim processes through carefully designed cache eviction sets. L1 data cache attacks are widely exploited and pose a significant privacy and confidentiality threat. Existing hardware-based countermeasures mainly focus on cache partitioning, randomization, and cache line flushing, which unfortunately either incur high overhead or can be circumvented by sophisticated attacks. In this paper, we propose a novel hardware-software co-design called BackCache with the idea of always achieving cache hits instead of cache misses to mitigate contention-based cache timing attacks on the L1 data cache. BackCache places the evicted cache lines from the L1 data cache into a fully associative backup cache to hide the evictions. To improve the security of BackCache, we introduce a randomly used replacement policy (RURP) and a dynamic backup cache resizing mechanism. Our evaluation on the gem5 simulator shows that BackCache can degrade the performance by 2.61%, 2.66%, and 3.36% For OS kernel, single-thread, and multi-thread benchmarks. Copyright © 2023, The Authors. All rights reserved.

关键词： Timing circuits

Fine-Grained Scene Graph Generation with Data Transfer

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Zhang, Ao Yao, Yuan Chen, Qianyu Ji, Wei Liu, Zhiyuan Sun, Maosong Chua, Tat-Seng Sea-NExT Joint Lab Singapore School of Computing National University of Singapore Singapore Department of Computer Science Technology Institute for Artificial Intelligence Tsinghua University Beijing China Beijing National Research Center for Information Science and Technology China

Scene graph generation (SGG) is designed to extract (subject, predicate, object) triplets in images. Recent works have made a steady progress on SGG, and provide useful tools for high-level vision and language understanding. However, due to the data distribution problems including long-tail distribution and semantic ambiguity, the predictions of current SGG models tend to collapse to several frequent but uninformative predicates (e.g., on, at), which limits practical application of these models in downstream tasks. To deal with the problems above, we propose a novel Internal and External Data Transfer (IETrans) method, which can be applied in a plug-and-play fashion and expanded to large SGG with 1,807 predicate classes. Our IETrans tries to relieve the data distribution problem by automatically creating an enhanced dataset that provides more sufficient and coherent annotations for all predicates. By applying our proposed method, a Neural Motif model doubles the macro performance for informative SGG. The code and data are publicly available at https://***/waxnkw/***. Copyright © 2022, The Authors. All rights reserved.

关键词： Data transfer

Numerical simulation of endovascular treatment options for cerebral aneurysms

学校读者我要写书评

暂无评论

GAMM Mitteilungen 2024年第3期47卷

作者： Frank, Martin Holzberger, Fabian Horvat, Medeea Kirschke, Jan Mayr, Matthias Muhr, Markus Nebulishvili, Natalia Popp, Alexander Schwarting, Julian Wohlmuth, Barbara University of the Bundeswehr Munich Institute for Mathematics and Computer-Based Simulation Neubiberg Germany Technical University of Munich School of Computation Information and Technology Department of Mathematics Chair for Numerical Mathematics Garching Germany Department of Neuroradiology Technical University of Munich Munich Germany University of the Bundeswehr Munich Data Science & Computing Lab Neubiberg Germany

Predicting the long-term success of endovascular interventions in the clinical management of cerebral aneurysms requires detailed insight into the patient-specific physiological conditions. In this work, we not only propose numerical representations of endovascular medical devices such as coils, flow diverters or Woven EndoBridge but also outline numerical models for the prediction of blood flow patterns in the aneurysm cavity right after a surgical intervention. Detailed knowledge about the postsurgical state then lays the basis to assess the chances of a stable occlusion of the aneurysm required for a long-term treatment success. To this end, we propose mathematical and mechanical models of endovascular medical devices made out of thin metal wires. These can then be used for fully resolved flow simulations of the postsurgical blood flow, which in this work will be performed by means of a Lattice Boltzmann method applied to the incompressible Navier–Stokes equations and patient-specific geometries. To probe the suitability of homogenized models, we also investigate poro-elastic models to represent such medical devices. In particular, we examine the validity of this modeling approach for flow diverter placement across the opening of the aneurysm cavity. For both approaches, physiologically meaningful boundary conditions are provided from reduced-order models of the vascular system. The present study demonstrates our capabilities to predict the postsurgical state and lays a solid foundation to tackle the prediction of thrombus formation and, thus, the aneurysm occlusion in a next step. © 2024 The Authors. GAMM - Mitteilungen published by Wiley-VCH GmbH.

关键词： Forecasting