检索结果-内蒙古大学图书馆

48th IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2023

作者： Tan, Huibin Hu, Kun Cao, Mingyu Wang, Mengzhu Xu, Liyang Yang, Wenjing Institute for Quantum Information State Key Laboratory of High Performance Computing China National University of Defense Technology College of Computer Science and Technology Changsha China

ISBN: (纸本)9781728163277

Tensor decomposition and reconstruction attention is a promising global context learning approach because it can remain efficient while avoiding feature compression. To exploit its potential even further in visual tracking, we redesign a 3D tensor modeling paradigm, namely tensor Decomposition, Interaction, Reconstruction attention (DIR), respectively corresponding to three function components, Tensor Decomposition Module (TDM), Tensor Interaction Module (TIM) and Context Reconstruction Module (CRM). Specifically, TDM decomposes a 3D tensor feature into rank-1 context fragments in different dimension views. The ingenuity here lies in the introduction of Circular Convolution for processing features at arbitrary scales and channel-sharing segments to enhance the interaction of the two branches in the Siamese network architecture. TIM obtains the tensor planes of each dimension by the Cross-Similarity operation of rank-1 tensors and fused cubic features, which brings more interactions between all feature dimensions. CRM reconstructs 3D context representations with the outputs of the above modules. In experiments, DIR is embedded into the tracker to verify its effectiveness. © 2023 IEEE.

关键词： context reconstruction Siamese tracking single object tracking tensor plane

来源：评论

学校读者我要写书评

暂无评论

Integrating Velox into TinkerPop for Graph Queries on Vectorized Engine 3

Integrating Velox into TinkerPop for Graph Queries on Vector...

引用

3rd International Conference on Electronic Information Engineering and computer, EIECT 2023

作者： Li, Zihao Xu, Liyang Jin, Ruochun Chen, Huan Tang, Yuhua National University of Defense Technology Institute for Quantum Information & State Key Laboratory of High Performance Computing College of Computer Science and Technology Changsha410073 China

ISBN: (纸本)9798350357707

To enhance the query efficiency of relational databases and build a unified computing backend, Meta has developed Velox, a vectorized execution engine library based on columnar storage, Currently, there is no standardized specification for computation engine, and storage in graph databases, leading to failed to effectively utilize the vectorized processing capability of modern CPU. In this paper, we propose a middleware that primarily focuses on (1) non-invasively integrating Velox into the TinkerPop framework to provide unified vectorized engine acceleration for all graph databases supporting the TinkerPop specification;(2) conducting graph queries based on the relational data storage model, eliminating the overhead of transforming the storage model into a graph storage model;(3) validating the acceleration effect of the vectorized engine on interactive workload of graph queries under a single-node environment. © 2023 IEEE.

关键词： Engines

来源：评论

学校读者我要写书评

暂无评论

Deep Representation Learning for Electron Ionization Mass Spectra Retrieval

Deep Representation Learning for Electron Ionization Mass Sp...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Liu, Shibo Zhang, Xuan Jiao, Anlei Sun, Shiwei Dian, Longyang Cui, Xuefeng Shandong University School of Computer Science and Technology Qingdao China Chinese Academy of Sciences Key Lab of Intelligent Information Processing Institute of Computing Technology Beijing China Shandong University State Key Laboratory of Microbial Technology Qingdao China

ISBN: (纸本)9798350386226

The task of retrieving and analyzing mass spectra is indispensable for the identification of compounds in mass spectrometry (MS). This methodology is of critical importance as it enables researchers to correlate observed spectra with established databases, thereby precisely determining the chemical composition of samples. The primary challenges to its efficacy lie in optimizing the balance between retrieval accuracy and processing speed. Empirical studies have demonstrated that by converting mass spectra into embeddings via deep learning, it is possible to achieve both high accuracy and speed in retrieval. Nevertheless, there are complex challenges associated with employing deep learning for spectral embedding, particularly within the domain of electron ionization mass spectrometry (EI-MS). In this paper, we introduce a novel representation learning technique termed EI-MS2VEC for EI-MS retrieval. Our spectrum retrieval methodology surpasses current state-of-the-art techniques such as FastEI. For the in-silico library, we attain hit rate@1 and hit rate@10 of 43.6% and 84.5%, respectively, compared to FastEI's 36.7% and 80.4%. Moreover, our retrieval approach operates with an order of magnitude greater speed than FastEI. The source code is available on Github (https://***/xfcui/EI-MS2VEC). © 2024 IEEE.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

SEER: A Time Prediction Model for CNNs from GPU Kernel's View 21

SEER: A Time Prediction Model for CNNs from GPU Kernel's Vie...

引用

Proceedings of the 30th International Conference on Parallel architectures and Compilation Techniques

作者： Guodong Liu Sa Wang Yungang Bao State Key Laboratory of Computer Architecture Institute of Computing Technology Beijing China and University of Chinese Academy of Sciences Beijing China State Key Laboratory of Computer Architecture Institute of Computing Technology Beijing China University of Chinese Academy of Sciences Beijing China and Institute of Computing Technology (Nanjing) Chinese Academy of Sciences Nanjing China State Key Laboratory of Computer Architecture Institute of Computing Technology Beijing China University of Chinese Academy of Sciences Beijing China and Peng Cheng Laboratory Shenzhen China

ISBN: (纸本)9781665442787

With the deepening of research and increasing size of data sets, deep neural networks have become larger and larger. To reduce the training time of large neural networks, researchers propose to optimize neural networks from different levels. When performing optimizations, prior knowledge about execution time of each part of the network can help avoid repeatedly time-consuming testing and profiling process. However it is quite challenging to build an accurate iteration time prediction model, due to opaque underlying implementation of network operators and complex architecture of accelerators. In this paper, we propose SEER, an iteration time prediction model for CNNs, targeting on GPU platforms. We propose to categorize convolution kernels into three different types: Compute-bound, DRAM-bound and Under-utilized, then we build performance model for each type respectively. We combined analytical models and learning-based models to make the performance model accurate and in line with GPU execution model. Experimental results show that, our model achieves 14.71% prediction error on convolution kernels and up to 1.79% prediction error for the overall computation time in one iteration of common CNNs. Besides, when used for selecting the best convolution algorithm, our model shows 7.14% lower error rate than cuDNN's official algorithm picker.

关键词： GPU

来源：评论

学校读者我要写书评

暂无评论

How to Train Your Neural Network for Molecular Structure Generation from Mass Spectra?

How to Train Your Neural Network for Molecular Structure Gen...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Zhao, Kai Liu, Yanmin Dian, Longyang Sun, Shiwei Cui, Xuefeng Shandong University School of Computer Science and Technology Qingdao China Shandong University State Key Laboratory of Microbial Technology Qingdao China Chinese Academy of Sciences Key Lab of Intelligent Information Processing Institute of Computing Technology Beijing China

ISBN: (纸本)9798350386226

Mass spectrometry serves as a pivotal tool for the analysis of small molecules through an examination of their mass-to-charge ratios. Recent advancements in deep learning have markedly enhanced the analysis of mass spectrometric data, facilitating the prediction of novel small molecule structures without the necessity of extensive databases. Nonetheless, the paucity of annotated datasets impedes the efficacious training of molecular generation models predicated on MS2 spectra. To mitigate this limitation, we introduce ctMSNovelist, an avant-garde method that amalgamates pre-training, fine-tuning, and co-training techniques to construct a more precise model for the generation of molecular structures from tandem mass spectrometry data. This novel approach augments both the training regimen and the predictive accuracy of the MSNovelist model, thereby surmounting the obstacle of limited data. The methodology commences with the pretraining of a Variational Autoencoder (VAE) to generate molecular fingerprints derived from SMILES strings. Subsequently, it undergoes fine-tuning to emulate noisy fingerprints originating from mass spectrometry (MS) data. Concurrently, MSNovelist is co-trained utilizing these simulated fingerprints, inclusive of the highly noisy variants produced in the early stages of VAE training. The incorporation of a substantial volume of noisy data serves to enhance model accuracy and avert overfitting. We evaluated ctMSNovelist using the GNPS dataset and attained a SMILES prediction accuracy of 48.8%, representing a 4.1% enhancement over MSNovelist. It is pertinent to note that the sole distinction between ctMSNovelist and MSNovelist in this experiment was the training process. The code and models are publicly available at https://***/xfcui/ctMSNovelist. © 2024 IEEE.

关键词： Neural networks

来源：评论

学校读者我要写书评

暂无评论

Monotonicity and nonmonotonicity in L3-valued propositional logic

引用

Frontiers of computer Science 2022年第4期16卷 33-43页

作者： Wei Li Yuefei Sui State Key Laboratory of Software Development Environment Beihang UniversityBeijing100083China Key Laboratory of Intelligent Information Processing Institute of Computing TechnologyChinese Academy of SciencesBeijing100190China School of Computer Science and Technology University of Chinese Academy of SciencesBeijing100049China

A sequent is a pair (Γ, Δ), which is true under an assignment if either some formula in Γ is false, or some formula in Δ is true. In L_(3)-valued propositional logic, a multisequent is a triple Δ∣Θ∣Γ, which is true under an assignment if either some formula in Δ has truth-value t, or some formula in Θ has truth-value m, or some formula in Γ has truth-value f. There is a sound, complete and monotonic Gentzen deduction system G for sequents. Dually, there is a sound, complete and nonmonotonic Gentzen deduction system G′ for co-sequents Δ: Θ: Γ. By taking different quantifiers some or every, there are 8 kinds of definitions of validity of multisequent Δ∣Θ∣Γ and 8 kinds of definitions of validity of co-multisequent Δ: Θ: Γ, and correspondingly there are 8 sound and complete Gentzen deduction systems for sequents and 8 sound and complete Gentzen deduction systems for co-sequents. Correspondingly their monotonicity is discussed.

关键词： sequent mulisequent gentzen deduction system monotonicity nonmonotonicity

来源：评论

学校读者我要写书评

暂无评论

Distributed Truss Computation in Dynamic Graphs

引用

Tsinghua Science and technology 2023年第5期28卷 873-887页

作者： Ziwei Mo Qi Luo Dongxiao Yu Hao Sheng Jiguo Yu Xiuzhen Cheng School of Computer Science and Technology Shandong UniversityQingdao 266200China State Key Laboratory of Software Development Environment School of Computer Science and Engineering and the Beijing Advanced Innovation Center for Big Data and Brain ComputingBeihang UniversityBeijing 100191China Big Data Institute Qilu University of Technology(Shandong Academy of Sciences)Jinan 250353China

Large-scale graphs usually exhibit global sparsity with local cohesiveness,and mining the representative cohesive subgraphs is a fundamental problem in graph *** k-truss is one of the most commonly studied cohesive subgraphs,in which each edge is formed in at least k 2 triangles.A critical issue in mining a k-truss lies in the computation of the trussness of each edge,which is the maximum value of k that an edge can be in a *** works mostly focus on truss computation in static graphs by sequential ***,the graphs are constantly changing dynamically in the real *** study distributed truss computation in dynamic graphs in this *** particular,we compute the trussness of edges based on the local nature of the k-truss in a synchronized node-centric distributed *** decomposing the trussness of edges by relying only on local topological information is possible with the proposed distributed decomposition ***,the distributed maintenance algorithm only needs to update a small amount of dynamic information to complete the *** experiments have been conducted to show the scalability and efficiency of the proposed algorithm.

关键词： distributed algorithm dynamic graph graph mining cohesive subgraph k-truss

来源：评论

学校读者我要写书评

暂无评论

Skyrmion transport driven by pure voltage generated strain gradient

引用

Chinese Physics B 2022年第11期31卷 566-570页

作者： Shan Qiu Jia-Hao Liu Ya-Bo Chen Yun-Ping Zhao Bo Wei Liang Fang Institute for Quantum Information&State Key Laboratory of High Performance Computing College of ComputerNational University of Defense TechnologyChangsha 410073China Hunan University of Humanities Science and Technology Loudi 417000China

The magnetic skyrmion transport driven by pure voltage-induced strain gradient is proposed and studied via micromagnetic *** combining the skyrmion with multiferroic heterojunction,a voltage-induced uniaxial strain gradient is adjusted to move *** the system,a pair of short-circuited trapezoidal top electrodes can generate the symmetric *** to the symmetry of strain,the magnetic skyrmion can be driven with a linear motion in the middle of the nanostrip without *** calculate the strain distribution generated by the trapezoidal top electrodes pair,and further investigate the influence of the strain intensity as well as the strain gradient on the skyrmion *** findings provide a stable and low-energy regulation method for skyrmion transport.

关键词： skyrmion strain gradient multiferroic heterojunction spintronics

来源：评论

学校读者我要写书评

暂无评论

NeuVSA: A Unified and Efficient Accelerator for Neural Vector Search 31

NeuVSA: A Unified and Efficient Accelerator for Neural Vecto...

引用

31st IEEE International Symposium on High Performance computer architecture, HPCA 2025

作者： Yuan, Ziming Dai, Lei Li, Wen Zhang, Jie Liang, Shengwen Wang, Ying Liu, Cheng Li, Huawei Li, Xiaowei Guo, Jiafeng Wang, Peng Chen, Renhai Zhang, Gong Shanxi University School of Computer and Information Technology China Peking University School of Computer Science China Institute of Computing Technology CAS State Key Lab of Processors China University of Chinese Academy of Sciences China Zhongguancun National Laboratory Beijing China Institute of Computing Technology CAS Key Lab of Network Data Science and Technology China Huawei Technologies Co. Ltd. China

ISBN: (纸本)9798331506476

Neural Vector Search (NVS) has exhibited superior search quality over traditional key-based strategies for information retrieval tasks. An effective NVS architecture requires high recall, low latency, and high throughput to enhance user experience and cost-efficiency. However, implementing NVS on existing neural network accelerators and vector search accelerators is sub-optimal due to the separation between the embedding stage and vector search stage at both algorithm and architecture levels. Fortunately, we unveil that Product Quantization (PQ) opens up an opportunity to break separation. However, existing PQ algorithms and accelerators still focus on either the embedding stage or the vector search stage, rather than both simultaneously. Simply combining existing solutions still follows the beaten track of separation and suffers from insufficient parallelization, frequent data access conflicts, and the absence of scheduling, thus failing to reach optimal recall, latency, and throughput. To this end, we propose a unified and efficient NVS accelerator dubbed NeuVSA based on algorithm and architecture co-design philosophy. Specifically, on the algorithm level, we propose a learned PQ-based unified NVS algorithm that consolidates two separate stages into the same computing and memory access paradigm. It integrates an end-to-end joint training strategy to learn the optimal codebook and index for enhanced recall and reduced PQ complexity, thus achieving smoother acceleration. On the architecture level, we customize a homogeneous NVS accelerator based on the unified NVS algorithm. Each sub-accelerator is optimized to exploit all parallelism exposed by unified NVS, incorporating a structured index assignment strategy and an elastic on-chip buffer to alleviate buffer conflicts for reduced latency. All sub-accelerators are coordinated using a hardware-aware scheduling strategy for boosted throughput. Experimental results show that the joint training strategy improves recall

关键词： Network embeddings

来源：评论

学校读者我要写书评

暂无评论

Formal Verification Based Synthesis for Behavior Trees 9th

Formal Verification Based Synthesis for Behavior Trees

引用

9th International Symposium on Dependable Software Engineering: Theories, Tools and Applications, SETTA 2023

作者： Hong, Weijiang Chen, Zhenbang Li, Minglong Li, Yuhan Huang, Peishan Wang, Ji College of Computer National University of Defense Technology Changsha China Key Laboratory of Software Engineering for Complex Systems National University of Defense Technology Changsha China Institute for Quantum Information & State Key Laboratory of High Performance Computing National University of Defense Technology Changsha China

ISBN: (纸本)9789819986637

Behavior trees (BTs) have been extensively applied in the area of both computer games and robotics, as the control architectures. However, the construction of BTs is labor-expensive, time-consuming, and even impossible as the complexity of task increases. In this work, we propose a formal verification based synthesis method to automatically construct BTs whose behaviors satisfy the given Linear Temporal Logic (LTL) specifications. Our method first explores candidate BTs by a grammar-based Monte Carlo Tree Search (MCTS), then the explored BTs are transformed into Communicating Sequential Processes (CSP) models. After that, we invoke the verifier to check the models’ correctness w.r.t. specifications, and provide feedback based on the verification result for guiding the search process. The application of our method on several representative robotic missions indicates its promising. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Behavior Trees CSP MCTS Synthesis

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：