检索结果-内蒙古大学图书馆

您好，读者！请登录

内蒙古大学图书馆

首页
概况
党建
资源
服务
科研支持
- 论文收录引用证明
- 科技查新
知识产权
档案馆
帮助

咨询与建议

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

您的常用邮箱：*

您的手机号码：*

问题描述：

当前已输入0个字，您还可以输入200个字

全部搜索
期刊论文
图书
学位论文
标准
纸本馆藏
外文资源发现
数据库导航
超星发现

高级检索

时间限定

出版年份：

文献类型

图书期刊文献学位论文多媒体

馆藏选择

电子馆藏纸本馆藏

核心期刊

全部期刊 SCI 收录期刊 SSCI 收录期刊 EI 收录期刊 CSCD 收录期刊 CSSCI 收录期刊

语言

中文英文

文献类型

期刊文献图书学位论文标准纸本馆藏

帮助

文字说明：

T=题名（书名、题名），A=作者（责任者），K=主题词，P=出版物名称，PU=出版社名称，O=机构（作者单位、学位授予单位、专利申请人），L=中图分类号，C=学科分类号，U=全部字段，Y=年（出版发行年、学位年度、标准发布年）

检索规则说明：

AND代表“并且”；OR代表“或者”；NOT代表“不包含”；(注意必须大写,运算符两边需空一格)

检索范例：

范例一：(K=图书馆学 OR K=情报学) AND A=范并思 AND Y=1982-2016
范例二：P=计算机应用与软件 AND (U=C++ OR U=Basic) NOT K=Visual AND Y=2011-2016

分类表

所选分类

>> <<

限定检索结果

文献类型

4,730 篇 会议
650 篇 期刊文献
3 篇 学位论文

馆藏范围

5,383 篇 电子文献
0 种 纸本馆藏

日期分布

学科分类号

1,500 篇 工学
- 835 篇 电气工程
- 774 篇 计算机科学与技术...
- 322 篇 信息与通信工程
- 239 篇 控制科学与工程
- 160 篇 软件工程
- 126 篇 机械工程
- 95 篇 仪器科学与技术
- 40 篇 电子科学与技术（可...
- 31 篇 交通运输工程
- 15 篇 航空宇航科学与技...
- 11 篇 测绘科学与技术
- 11 篇 生物医学工程（可授...
- 10 篇 土木工程
- 9 篇 网络空间安全
- 7 篇 材料科学与工程（可...
- 4 篇 动力工程及工程热...
- 4 篇 矿业工程
- 4 篇 石油与天然气工程
166 篇 理学
- 68 篇 系统科学
- 56 篇 数学
- 20 篇 物理学
- 19 篇 生物学
- 9 篇 统计学（可授理学、...
- 5 篇 地球物理学
- 4 篇 地理学
100 篇 管理学
- 98 篇 管理科学与工程(可...
49 篇 医学
- 45 篇 临床医学
- 4 篇 基础医学(可授医学...
38 篇 文学
- 37 篇 新闻传播学
7 篇 艺术学
- 7 篇 设计学（可授艺术学...
6 篇 农学
2 篇 教育学
1 篇 军事学

主题

5,383 篇 inference algori...
540 篇 computational mo...
408 篇 fuzzy systems
400 篇 algorithm design...
386 篇 fuzzy logic
380 篇 bayesian methods
370 篇 computer science
343 篇 training
333 篇 neural networks
304 篇 approximation al...
291 篇 bayes methods
236 篇 optimization
225 篇 clustering algor...
219 篇 signal processin...
211 篇 data mining
207 篇 fuzzy sets
206 篇 accuracy
205 篇 heuristic algori...
192 篇 prediction algor...
185 篇 feature extracti...

机构

15 篇 school of electr...
11 篇 university of ch...
10 篇 school of comput...
10 篇 zhejiang univ co...
9 篇 tsinghua univers...
9 篇 purple mt labs p...
9 篇 national univers...
8 篇 microsoft resear...
8 篇 university of mi...
7 篇 department of co...
7 篇 tianjin univ col...
7 篇 department of el...
7 篇 university of so...
7 篇 department of au...
7 篇 tsinghua univ de...
6 篇 ibm thomas j. wa...
6 篇 nanyang technol ...
6 篇 university of so...
6 篇 department of el...
6 篇 department of co...

作者

8 篇 pushmeet kohli
7 篇 li cheng
7 篇 edward r. doughe...
7 篇 wei li
7 篇 xiaodong wang
7 篇 t.s. huang
7 篇 erik b. sudderth
7 篇 a. taylan cemgil
7 篇 p. baranyi
7 篇 nizar bouguila
6 篇 simon godsill
6 篇 alan s. willsky
6 篇 hui wang
6 篇 wei liu
6 篇 ying wu
6 篇 wang wei
6 篇 s. suresh
6 篇 chen yongxin
6 篇 yang xu
6 篇 luca benini

语言

5,319 篇 英文
51 篇 其他
11 篇 中文
1 篇 西班牙文

检索条件"主题词=Inference algorithms"

共 5383 条记录，以下是171-180 订阅

全选清除本页清除全部题录导出标记到"检索档案"

详细简洁

排序：

EVASION: Efficient KV CAche CompreSsion vIa PrOduct QuaNtization

EVASION: Efficient KV CAche CompreSsion vIa PrOduct QuaNtiza...

引用

Design, Automation and Test in Europe Conference and Exhibition

作者： Zongwu Wang Fangxin Liu Peng Xu Qingxiao Sun Junping Zhao Li Jiang Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai Qi Zhi Institute Dept. of CST SSSLab China University of Petroleum-Beijing China Ant Group MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University

ISBN: (数字)9783982674100

ISBN: (纸本)9798331534646

Large language models (LLMs) are increasingly utilized for complex tasks requiring longer context lengths, with some models supporting up to 128K or 1M tokens. This trend, however, presents significant challenges in inference speed and memory management. The primary bottleneck in long-context LLM inference is the quadratic computational complexity of attention mechanisms, causing substantial slowdowns as sequence length increases. KV cache mechanism alleviates this issue by storing pre-computed data, but introduces memory requirements that scale linearly with context length, hindering efficient LLM deployment. Quantization emerges as a promising approach to address the widening gap between LLM size and memory capacity. However, traditional quantization schemes often yield suboptimal compression results for KV caches due to two key factors: i) On-the-fly quantization and de-quantization, causing significant performance overhead; ii) Prevalence of outliers in KV values, challenging low-bitwidth uniform quantization. To this end, we propose EVASION, a novel quantization framework achieving low-bitwidth KV cache through product quantization. First, we conduct a thorough analysis of KV cache distribution, revealing the limitations of existing quantization schemes. Second, we introduce a non-uniform quantization algorithm based on product quantization, which efficiently compresses data while preserving accuracy. Third, we develop a high-performance GPU inference framework for EVASION that leverages sparse computation and asynchronous quantization, significantly enhancing inference speed. Comprehensive evaluation results demonstrate that EVASION can achieve 4 bits quantization trivial perplexity and accuracy loss.

关键词： Quantization (signal) Accuracy Large language models Memory management Graphics processing units Europe Market research inference algorithms Low latency communication Context modeling

来源：评论

学校读者我要写书评

暂无评论

Enhancing Low-Resolution Person Attribute Recognition Through the Integration of Super Resolution and Attribute Learning

Enhancing Low-Resolution Person Attribute Recognition Throug...

引用

Women in Data Science at Prince Sultan University (WiDS PSU), International Conference of

作者： Ramin Abbaszadi Niayesh Gharaei Department of Software Engineering Faculty of Engineering OSTİM Technical University Ankara Turkey Department of Computer Engineering Faculty of Engineering OSTİM Technical University Ankara Turkey

ISBN: (数字)9798331520922

ISBN: (纸本)9798331520939

In the context of video monitoring or surveillance, particularly in safe city initiatives, visual attributes of individuals such as gender, backpacks, and types of clothing are essential for person search and re-identification. Effectively detecting and extracting these attributes usually necessitates high-quality videos and images, which are often lacking in standard surveillance footage. Beyond improving hardware technology, enhancing inference algorithms for low-resolution (LR) data is essential. This research presents two solutions: developing a unified neural network architecture derived from existing models and introducing a new method for re-identification in LR videos. The proposed architecture, named SRMAR, incorporates Super Resolution (SR) alongside Multi-Attribute Recognition (MAR) models into a single neural network, thereby improving recognition effectiveness. Tests conducted on two benchmark datasets show that the architecture is effective and suitable for recognizing multiple attributes in LR images.

关键词： Visualization Image recognition Surveillance Superresolution Neural networks Urban areas Benchmark testing Network architecture inference algorithms Standards

来源：评论

学校读者我要写书评

暂无评论

Machine Learning-Driven Binocular Vision Optimization for Enhanced 3D Reconstruction of IP Images

Machine Learning-Driven Binocular Vision Optimization for En...

引用

Machine Learning and Autonomous Systems (ICMLAS), International Conference on

作者： Zhouyang Dongfang Chen Hang Xiamen Institute of Technology School of Arts and Communication Xiamen China LUOHE VOCATIONAL TECHNOLOGY COLLEGE HENAN China

ISBN: (数字)9798331505745

ISBN: (纸本)9798331505752

As an important branch of machine learning, 3D reconstruction algorithms are widely used in visual reconstruction tasks. To overcome the limitations of traditional 3D reconstruction methods in terms of parameter set, inference speed and accuracy, this study proposes a new 3D reconstruction model based on binocular vision optimization algorithm. As a specific scenario, the algorithm is applied to the processing of complex scenes in IP images. The proposed model first combines the prior knowledge of image shape and models it by elliptical feature point sampling and grid multi-homography estimation. This operation can achieve efficient disparity computation; then, the study uses deep learning to optimize the binocular matching algorithm. The algorithm accurately predicts the disparity value through the stacked cost volume construction technology. At the same time, as an algorithm optimization, this study optimizes the reconstruction algorithm based on the optimization strategy of PatchMatch and further reduces the computational complexity. The experiment uses a self-constructed image dataset to verify the performance of the model in various scenarios. The experiment shows that the proposed algorithm has significant advantages in the accuracy and efficiency of 3D reconstruction.

关键词： Solid modeling Three-dimensional displays Machine learning algorithms Accuracy Shape Computational modeling Prediction algorithms inference algorithms IP networks Optimization

来源：评论

学校读者我要写书评

暂无评论

Robust Hybrid Convolutional Beamspace for High Resolution DOA Estimation

Robust Hybrid Convolutional Beamspace for High Resolution DO...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Rubin Jose Peter Sooraj K. Ambat Naval Physical & Oceanographic Laboratory (NPOL) Defence R&D Organization (DRDO) Kochi India

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Hybrid beamforming realizes a combination of analog and digital processing of sensor data using a reduced number of digital channels. The analog combiner in the analog stage decreases the number of analog-to-digital converters (ADCs), resulting in minimized hardware cost and power consumption. Recently, a hybrid convolutional beamspace (CBS) beamformer was proposed to achieve lower computational complexity, higher direction-of-arrival (DOA) resolution, and a smaller mean-squared-error (MSE). However, the hybrid CBS method assumes a well-calibrated uniform linear array (ULA) for DOA estimation, which is vulnerable to gain and phase mismatches in the sensor elements that degrade the performance of the hybrid beamformer. To address this limitation, in this paper, a robust hybrid convolutional beamspace method is proposed for DOA estimation, where calibration and high-resolution estimation of source directions are carried out simultaneously in the beamspace using variational inference.

关键词： Phased arrays Direction-of-arrival estimation Power demand Convolution Array signal processing Estimation Signal processing algorithms inference algorithms Speech processing Sensor arrays

来源：评论

学校读者我要写书评

暂无评论

Adaptive Agriculture System with Soil Health Prediction using Ensemble Techniques and Smart Irrigation using Deep Learning

Adaptive Agriculture System with Soil Health Prediction usin...

引用

Advancement in Computation & Computer Technologies (InCACCT), International Conference on

作者： Vasan J K Rithvidas R Hariprasad S M Maria Dominic Savio R S Haritha Sudeshna Shettygari Department of Electronics and Communication Engineering SRM Institute of Science and Technology Chennai India Department of Computational Intelligence School of Computing SRM Institute of Science and Technology Chennai India

ISBN: (数字)9798331543891

ISBN: (纸本)9798331543907

Agriculture is a cornerstone of the country’s economy, serving as a fundamental necessity for human survival and a critical driver of socio-economic stability. In traditional farming, the efficiency of crop yield is often limited, and resources are not managed effectively, leading to a significant reduction in overall productivity. This paper introduces an IoT-based adaptive agriculture system using a real-time soil health prediction algorithm, where it leverages the Custom Stacked Ensemble model combined with K-Fold Stratified cross-validation to generate predictions from dynamic environmental and soil conditions. The proposed model achieves an accuracy of 94.5%. Additionally, the system incorporates a smart irrigation mechanism that regulates water delivery to crops, ensuring optimal growth while conserving water resources. The irrigation process is automated using the Enhanced GRU model, which achieves an accuracy of 99.43%, further enhancing efficiency and sustainability. The system collects the data from the sensor in the field via a LoRa-based network that communicates with each other to transmit the data to the IoT server. By combining the weather forecasts and the data obtained from the sensors, the system gives soil suitability for the farmer and also predicts the precise irrigation levels and soil fertility, aiding adaptive farming.

关键词： Irrigation Adaptation models Adaptive systems Accuracy Weather forecasting Soil Prediction algorithms inference algorithms Water resources Farming

来源：评论

学校读者我要写书评

暂无评论

Lightweight Convolutional YOLOv5 for Real-Time Object Detection in Infrared Stereo Images

Lightweight Convolutional YOLOv5 for Real-Time Object Detect...

引用

2025 2nd International Conference on algorithms, Software Engineering and Network Security (ASENS)

作者： Weishen Li Yuqing Lao Xiao Zhong Jianya Wang Wenxu Zhao Hao Tang School of Transportation and Economic Management Guangdong Communication Polytechnic Guangzhou China School of International Education Yellow River Conservancy Technical Institute Kaifeng China

ISBN: (数字)9798331520298

ISBN: (纸本)9798331520304

Since under low-illumination and high-interference environments, infrared binocular images still have more accurate depth and target information, how to effectively achieve real-time and accurate object detection on the embedded platform. In this research, we propose an improved YOLOv5-based model which is the fusion of a lightweight convolution backbone network and a multi-scale attention fusion-detection head, which is composed of two parts: a feature extraction stage is designed to introduce separable convolution and group normalization to reduce the number of parameters and accelerate the inference speed; and in the detection stage, a deep-learnable dual-channel attention module (channel attention + spatial attention) is used to fuse and weight the multi-scale features of infrared left and right eye images. At last, with the high detection accuracy, the model complexity and inference delay were greatly decreased, which is applicable to the application scenarios with high requirement for system resources and real-time performance. Experimental results demonstrate that the proposed lightweight convolutional YOLOv5 model can strike a balance between excellent accuracy and real-time performance on resource-constrained platforms for infrared binocular image object detection.

关键词： YOLO Accuracy Head Convolution Software algorithms Network security Feature extraction Real-time systems inference algorithms Software engineering

来源：评论

学校读者我要写书评

暂无评论

Efficient Pothole Detection in Video Sequences Using YOLOv4 with Cosine Similarity: A Road Safety Solution

Efficient Pothole Detection in Video Sequences Using YOLOv4 ...

引用

Intelligent and Innovative Technologies in Computing, Electrical and Electronics (IITCEE), International Conference

作者： Anish Kumar R Mohit Sai Reddy Sharath B Laxmi V Mona Department of ISE BNMIT Department of CSE BNMIT

ISBN: (数字)9798331515911

ISBN: (纸本)9798331515928

A novel approach for the detection and counting of potholes in road surfaces using video sequences is proposed in this paper. Potholes are a significant road hazard that can lead to accidents and damage to vehicles. Accurate and efficient detection and counting of potholes are crucial for road maintenance and safety. State-of-the-art deep learning methods, including Convolutional Neural Networks (CNN), Single Shot MultiBox Detector (SSD), You Only Look Once version 4 (YOLOv4), and YOLOv4-tiny are used for pothole detection in video sequences. Additionally, a counting method based on cosine similarity and tracker to estimate the total number of potholes is proposed. The experimental results demonstrate the effectiveness of the proposed methods. These results highlight that YOLOv4 is better in accurate detection of potholes in video sequences. Moreover, the proposed counting method based on cosine similarity achieves the best standard error rate, indicating its capability to estimate the number of potholes with reasonable accuracy. This research provides a valuable contribution to the field of road maintenance and safety by offering an efficient solution for pothole detection and counting, which can aid authorities in timely road repairs and enhance road users' safety.

关键词： YOLO Deep learning Accuracy Tracking Roads Video sequences Prediction algorithms inference algorithms Maintenance Standards

来源：评论

学校读者我要写书评

暂无评论

Panacea: Novel DNN Accelerator using Accuracy-Preserving Asymmetric Quantization and Energy-Saving Bit-Slice Sparsity

Panacea: Novel DNN Accelerator using Accuracy-Preserving Asy...

引用

IEEE Symposium on High-Performance Computer Architecture

作者： Dongyun Kam Myeongji Yun Sunwoo Yoo Seungwoo Hong Zhengya Zhang Youngjoo Lee Pohang University of Science and Technology (POSTECH) University of Michigan

ISBN: (数字)9798331506476

ISBN: (纸本)9798331506483

Low bit-precisions and their bit-slice sparsity have recently been studied to accelerate general matrix-multiplications (GEMM) during large-scale deep neural network (DNN) inferences. While the conventional symmetric quantization facilitates low-resolution processing with bit-slice sparsity for both weight and activation, its accuracy loss caused by the activation’s asymmetric distributions cannot be acceptable, especially for largescale DNNs. In efforts to mitigate this accuracy loss, recent studies have actively utilized asymmetric quantization for activations without requiring additional operations. However, the cuttingedge asymmetric quantization produces numerous nonzero slices that cannot be compressed and skipped by recent bit-slice GEMM accelerators, naturally consuming more processing energy to handle the quantized DNN *** simultaneously achieve high accuracy and hardware efficiency for large-scale DNN inferences, this paper proposes an Asymmetrically-Quantized bit-Slice GEMM (AQS-GEMM) for the first time. In contrast to the previous bit-slice computing, which only skips operations of zero slices, the AQS-GEMM compresses frequent nonzero slices, generated by asymmetric quantization, and skips their operations. To increase the slicelevel sparsity of activations, we also introduce two algorithm-hardware co-optimization methods: a zero-point manipulation and a distribution-based bit-slicing. To support the proposed AQS-GEMM and optimizations at the hardware-level, we newly introduce a DNN accelerator, Panacea, which efficiently handles sparse/dense workloads of the tiled AQS-GEMM to increase data reuse and utilization. Panacea supports a specialized dataflow and run-length encoding to maximize data reuse and minimize external memory accesses, significantly improving its hardware efficiency. Numerous benchmark evaluations show that Panacea outperforms existing DNN accelerators, e.g., $1.97 \times$ and $3.26 \times$ higher energy efficiency, and $1.88 \tim

关键词： Quantization (signal) Accuracy Symmetric matrices Single instruction multiple data Memory management Artificial neural networks Throughput Hardware inference algorithms Optimization

来源：评论

学校读者我要写书评

暂无评论

ReVerie: Contemplating Agency and Truth in Al-assisted Immersive Dream Reliving

ReVerie: Contemplating Agency and Truth in Al-assisted Immer...

引用

Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW), IEEE Conference on

作者： Pinyao Liu Keon Ju M. Lee Independent Artist-Researcher School of Interactive Arts + Technology Simon Fraser University

ISBN: (数字)9798331514846

ISBN: (纸本)9798331525637

ReVerie AI installation explores the scientific concept of dream reliving. Visitors engage in a ritualistic act of whispering their dreams. Each dream transcription and corresponding visualization are temporarily stored in the memory of the AI system. After that, the ReVerie AI system makes an inference from this memory to generate a shared, evolving A/V experience, where participants fly through interconnected dream objects in a virtual space. At its core, ReVerie challenges the capacity of technology to authentically represent the intangible and private terrain of dreams. By inviting participants to contribute whispered dreams that are algorithmically transformed into 3D dream objects, ReVerie pushes them to confront whether such digitally synthesized representations can capture the true essence of our subconscious—or if they craft new forms of detachment or control over personal narratives instead.

关键词： Visualization Three-dimensional displays Art Generative AI Conferences Virtual reality User interfaces inference algorithms

来源：评论

学校读者我要写书评

暂无评论

An Improved Serial IMPLY Adder Algorithm for Efficient Neural Network Applications

An Improved Serial IMPLY Adder Algorithm for Efficient Neura...

引用

IEEE Latin American Symposium on Circuits and Systems (LASCAS)

作者： Fabian Seiler Nima TaheriNejad Technische Universität Wien (TU Wien) Austria Heidelberg University Germany

ISBN: (数字)9798331522124

ISBN: (纸本)9798331522131

Memristive systems are one of the most promising candidates for a post-CMOS era. They are small, energy-efficient, and are ideal targets for In-Memory Computation (IMC) via stateful logic. As adders are critical building blocks for any computing systems, improving them is an essential design goal. With the rise of Artificial Intelligence (AI), providing memristive adders that are optimized for Neural Networks (NNs) is extremely important. For this, we propose a Material Implication (IMPLY)-based adder algorithm in the serial topology that can preserve the weights in memory, which was not addressed in the State-of-the-Art (SoA). Our approach is 20% - 23% faster and requires 1% - 12% less energy when the adder is used repeatedly. We propose a flowchart for IMPLY-based algorithms that can represent the state changes of individual memristors and apply it to our adder. We embed our adder in a shift-and-add multiplier and evaluate the potential gains on the 8-bit quantized ResNet18. Our approach is up to 17% more energy-efficient and requires up to 20% fewer cycles for the inference than SoA adder.

关键词： Flowcharts Visualization Network topology Memristors inference algorithms Energy efficiency Topology Logic Artificial intelligence Adders

来源：评论

学校读者我要写书评

暂无评论

没有更多数据了...

全选清除本页清除全部题录导出标记到“检索档案”

共500页 << < 14 15 16 17 18 19 20 21 22 23 > >>

检索报告对象比较合并检索0

隐藏清空

合并搜索

回到顶部

执行限定条件

内容：

评分：

请选择保存的检索档案：

请选择收藏分类：

订阅名称：

通借通还

温馨提示：

图书名称：

借书校区：

取书校区：

手机号码：

邮箱地址：

一卡通帐号：

电话和邮箱必须正确填写，我们会与您联系确认。

联系人：

所在院系：

联系邮箱：

联系电话：

内蒙古自治区呼和浩特市赛罕区大学西街235号邮编: 010021

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：