检索结果-内蒙古大学图书馆

7th International conference on Electronics Technology, ICET 2024

作者： Lu, Hui Zhang, Yi Wu, Zebin Wei, Zhihui Nanjing University of Science and Technology School of Computer Science and Engineering Nanjing China

ISBN: (纸本)9798350363951

During the last 3 years, researchers have endeavored to extend the efficacy of contrastive learning (CL) towards addressing the challenges inherent in cloud workload prediction, which has demonstrated considerable success in the domains of computer Vision (CV) and Natural Language Processing(NLP). However, due to the distinctive temporal characteristics of serialized workload information, relying solely on empirical guidance from other domains may prove insufficient for cloud workload prediction. Therefore, we systematically investigated three key components of CL, including: 1) designing backbone encoder for feature extraction, 2) designing methods for extracting negative samples for contrast, and 3) designing the CL loss. We found that inappropriate construction of negative samples may introduce excessive pseudo-negative samples, which neither preserve temporal characteristics nor provide sufficient discriminative features. Additionally, focusing solely on the discrimination between positive and negative samples in the CL loss may not be adequate for enabling the model to learn the temporal patterns of serialized workload information. To address these problems, we propose a novel self-supervised model named Cloudformer. Specifically, we first constructed a CL network based on the Transformer architecture using a pre-train-finetune framework. Following this, we proposed a method to filter pseudo-negative samples based on differential cosine similarity, aimed at assisting the model in more effectively distinguishing between positive and negative samples. Additionally, we introduced the BatchNCE loss, combining MSE loss with InfoNCE loss as a joint optimization objective for CL, effectively enhancing the predictive capability of the model. The experimental results indicate that Cloudformer is capable of learning high-quality patterns in serialized workload sequences and achieves more advanced performance in short sequence prediction tasks. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Editors' Notes

引用

computer MUSIC JOURNAL 2022年第1-2期46卷 5-7页

作者： Cadiz, Rodrigo F. Schumacher, Federico

Despite computer music being commonly identified as having North American or European origins, Latin American composers and performers have had a great interest in it since its inception, with a long and prolific history comparable to these other traditions. Research and music from this region have had limited visibility, however, resulting in a significant gap in academic knowledge, partially due to the language barrier and the lack of technological means and international exposure for many artists. Nevertheless, in recent times Latin America has experienced an interesting increase in activity in the field, as some of the most important international academic conferences have taken place in the region, such as the International conference on New Interfaces for Musical Expression (NIME) in Brazil in 2019, and both the International computer Music conference (ICMC) and the International conference on Live Coding (ICLC) in Chile in 2021. These conferences have showcased a significant amount of artistic work and research that is being accomplished in the region. This special issue ofcomputer Music Journalaims to bridge some of this still-existing gap by publishing new technical and historical research on computer music practice in Latin *** special issue comprises eight articles written by artists and researchers with ties to Argentina, Brazil, Chile, Mexico, and Uruguay, offering some fresh perspectives on the computer music that Latin America is currently producing and addressing some of the most important challenges that it faces *** begin with the article of Ricardo Dal Farra, who is an important Argentinian composer and researcher of the history of electroacoustic and computer music in Latin America. He argues that the progression of this genre in the region has been astounding despite political and economic instability that has affected the lives of inhabitants for decades. One example of this development is the Latin American Electroacoustic Music

关键词：

来源：评论

学校读者我要写书评

暂无评论

How to Leverage Diverse Demonstrations in Offline Imitation Learning 41

How to Leverage Diverse Demonstrations in Offline Imitation ...

引用

41st International conference on Machine Learning, ICML 2024

作者： Yue, Sheng Liu, Jiani Hua, Xingyuan Ren, Ju Lin, Sen Zhang, Junshan Zhang, Yaoxue Department of Computer Science and Technology Tsinghua University Beijing China Zhongguancun Laboratory Beijing China Department of Computer Science University of Houston Texas United States Department of Electrical and Computer Engineering University of California Davis United States

Offline Imitation Learning (IL) with imperfect demonstrations has garnered increasing attention owing to the scarcity of expert data in many real-world domains. A fundamental problem in this scenario is how to extract positive behaviors from noisy data. In general, current approaches to the problem select data building on state-action similarity to given expert demonstrations, neglecting precious information in (potentially abundant) diverse state-actions that deviate from expert ones. In this paper, we introduce a simple yet effective data selection method that identifies positive behaviors based on their resultant states - a more informative criterion enabling explicit utilization of dynamics information and effective extraction of both expert and beneficial diverse behaviors. Further, we devise a lightweight behavior cloning algorithm capable of leveraging the expert and selected data correctly. In the experiments, we evaluate our method on a suite of complex and high-dimensional offline IL benchmarks, including continuous-control and vision-based tasks. The results demonstrate that our method achieves state-of-the-art performance, outperforming existing methods on 20/21 benchmarks, typically by 2-5x, while maintaining a comparable runtime to Behavior Cloning (BC). Copyright 2024 by the author(s)

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Symmetric Locality: Definition and Initial Results

Symmetric Locality: Definition and Initial Results

引用

high performance Computing, Networking, Storage and Analysis, SC-W: Workshops of the International conference for

作者： Giordan Escalona Dylan McKellips Chen Ding Department of Computer Science University of Rochester Rochester NY

ISBN: (数字)9798350355543

ISBN: (纸本)9798350355550

In this paper, we characterize symmetric locality. In designing algorithms, compilers, and systems, data movement is a common bottleneck in high-performance computation, in which we improve cache and memory performance. We study a special type of data reuse in the form of repeated traversals, or re-traversals, which are based on the symmetric group. The cyclic and sawtooth traces are previously known results in symmetric locality, and in this work, we would like to generalize this result for any re-traversal. Then, we also provide an abstract framework for applications in compiler design and machine learning models to improve the memory performance of certain programs.

关键词： program processors high performance computing conferences Computational modeling Machine learning

来源：评论

学校读者我要写书评

暂无评论

Robot control with multitasking of brain-computer interface 17

Robot control with multitasking of brain-computer interface

引用

17th International conference on Control, Automation, Robotics and Vision (ICARCV)

作者： Zhou, Yajun Lu, Zilin Li, Yuanqing South China Univ Technol Sch Automat Sci & Engn Guangzhou 510640 Peoples R China Res Ctr Brain Comp Interface Pazhou Lab Guangzhou 510330 Peoples R China

ISBN: (纸本)9781665476874

Brain-computer interfaces (BCI) have been extensively researched to assist people with motor paralysis in controlling external devices such as a robotic limb. However, most BCI systems required participants to focus on a single task, limiting their ability to generate other mental or physical activities. Therefore, people's performance of the BCI-based robotic control in multitasking was discussed, as eight healthy subjects performed motor-related tasks of motor imagery and two-handed balancing ball movement, while simultaneously performing visuospatial attention to asynchronously trigger "drinking" actions of a humanoid robot arm with accuracies of 90% and 87.5%, respectively. The online results indicate that the BCI-based robot control system developed for multi-task conditions has a high potential for human augmentation.

关键词： Anthropomorphic robots

来源：评论

学校读者我要写书评

暂无评论

On the impact of pre-training datasets for matching dendritic identifiers using residual nets

On the impact of pre-training datasets for matching dendriti...

引用

2024 International Workshop on Artificial Intelligence for Signal, Image Processing and Multimedia, AI-SIPM 2024

作者： Jhaj, Baaz Shukla, Ankita Turaga, Pavan Kozicki, Michael Geometric Media Lab Arizona State University TempeAZ United States Computer Science and Engineering University of Nevada RenoNV United States School of Electrical Computer and Energy Engineering Arizona State University TempeAZ United States

ISBN: (纸本)9798400705489

Dendrites are easy to synthesize branching structures that exhibit randomness;yet they are unique, non-repeatable, and identifiable with the right algorithmic innovations. This has created a novel application area where manufactured dendritic structures are being used as product identifiers - essentially "fingerprints for things". Unlike barcodes, which are linear structures, dendrites exhibit spatial randomness. This, coupled with a unique optical signal generated by light scattering from material inhomogeneities, ensures that each dendrite is unique and unclonable. While there have not yet been any established methods on reading dendritic patterns for verification using image data, identifying dendrites using computer vision techniques could have high potential. Due to limited data and low variance, dendrite identification can be considered to be a fine-grained classification task. In this paper, we examine how the selection of pre-trained models influences dendrite classification. The dendrites we work with share similarity to human fingerprints, thus we begin with a model trained for matching fingerprint data to extract features relevant to dendrites. Additionally, we explore broader pre-training approaches, using ImageNet-1K for our second model and ImageNet-21K for our third model. Surprisingly, our results indicate that even with the visual similarity with human fingerprints, more general pre-training with common image datasets achieves better performance on dendrite classification. © 2024 ACM.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Impact of Processor Frequency Scaling on performance and Energy Consumption for WZ Factorization on Multicore Architecture

Impact of Processor Frequency Scaling on Performance and Ene...

引用

Federated conference on computer Science and Information Systems (FedCSIS)

作者： Beata Bylina Jaroslaw Bylina Monika Piekarz Institute of Computer Science Marie Curie-Sklodowska University Lublin Poland

With the growing demand for computing power, new multicore architectures have emerged to provide better performance. Reducing their energy consumption is one of the main challenges in achieving high performance computing. Current research trends develop new software and hardware techniques to achieve the best performance and energy compromise. In this work, we investigate the effect of processor frequency scaling using Dynamic Voltage Frequency Scaling on performance and energy consumption for the WZ factorization. This factorization is implemented both without optimization techniques and with strip mining. This technique involves transforming the program loop to improve program performance. Based on time and energy tests, we have shown that for the WZ factorization algorithm, regardless of the presence of manual optimization, it pays to reduce the frequency to save energy without losing performance. The conclusion can be extended to analogous algorithms — also having a high ratio of memory access to computational operations.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Towards Continual Federated Learning of Monocular Depth for Autonomous Vehicles 100

Towards Continual Federated Learning of Monocular Depth for ...

引用

100th IEEE Vehicular Technology conference, VTC 2024-Fall

作者： De S. Soares, Elton F. Campos, Carlos Alberto V. Ibm Research Brazil Research Lab RJ Rio de Janeiro20031-170 Brazil Federal University of the State of Rio de Janeiro Graduate Program in Informatics RJ Rio de Janeiro22290-255 Brazil

ISBN: (纸本)9798331517786

Recent investigations in computer vision for autonomous vehicles have focused on depth estimation from images, owing to its cost-efficiency and adaptability. Monocular depth estimation, using a single camera, is notable for its adaptability compared to binocular techniques that require two fixed cameras. Sophisticated methodologies employ self-supervised deep neural networks, while latest research proposes the use of federated learning to tackle crucial challenges for autonomous driving, such as data privacy, network usage, computation distribution, and connectivity robustness. Nevertheless, continual learning presents an additional challenge for both centralized and federated training, as updating models with novel datasets may induce the forgetting of previously acquired samples, thereby diminishing the model's versatility. Meanwhile, recent research indicates that experience replay strategies can be used to alleviate forgetting in autonomous driving use cases. We introduce ERFedSCDepth, an innovative approach amalgamating experience replay, federated learning, and deep self-supervision to allow the training of monocular depth estimators with high effectiveness and efficiency. Assessment using KITTI and DDAD datasets shows the efficacy of our approach, achieving enhanced continual learning performance over state-of-the-art baseline at standard depth metrics. © 2024 IEEE.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Learning to Distinguish Entities Overlapping for Open Relation Extraction 9

Learning to Distinguish Entities Overlapping for Open Relati...

引用

9th IEEE International conference on Data Science in Cyberspace, DSC 2024

作者： Li, ChangJian Chen, Kai Yu, Han Huang, YinXuan Zhao, XueChen Zhang, JianFeng Li, AiPing National University of Defense Technology College of Computer Changsha China

ISBN: (纸本)9798350391367

Open relation extraction (OpenRE) is a challenging task in natural language processing (NLP) which aims to identify and extract novel and previously unseen relationships from textual data without relying on predefined relation types. Traditional supervised methods for relation extraction (RE) require extensive manual annotations, which can be costly and limiting. To address this, we propose a novel approach, Learning to Distinguish Entities Overlapping (LDEO), which leverages metric learning to enhance the representation of relation features and effectively distinguish instances with overlapping entities. Our method introduces two data augmentation strategies to enrich the training dataset with positive and negative instances, facilitating the learning process, despite the scarcity of overlapping entity instances. Furthermore, we employ an autoencoder for dimensionality reduction to mitigate the challenges associated with high-dimensional feature spaces, such as increased computational complexity and data sparsity. Extensive experiments on two datasets demonstrate that our LDEO approach significantly outperforms state-of-the-art models, with an average improvement of 5% and 14.2% in performance metrics on FewRel and TACRED datasets, respectively. Our work contributes to advancing the field of OpenRE by providing a robust and efficient method for distinguishing overlapping entities and extracting open-domain relations. © 2024 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

PMemTrace: Lightweight and Efficient Memory Access Monitoring for Persistent Memory 1

引用

22nd International conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2022

作者： Zhang, Yushuqing Lu, Kai Wu, Zhenwei Zhang, Wenzhe College of Computer National University of Defense Technology Changsha China

ISBN: (数字)9783031226779

ISBN: (纸本)9783031226762

Persistent memory gains increasing popularity in recent years. Many persistent memory systems leverage memory access monitoring to achieve user-transparent crash consistency enforcement. However, traditional program analysis-oriented memory access monitoring mechanisms incur prohibitive performance overheads in persistent memory applications. Especially, for applications developed for systems with hybrid DRAM and persistent memory, existing memory tracking approaches cannot distinguish volatile and persistent memory writes, which leads to significant unnecessary runtime cost. This paper describes PMemTrace, a lightweight and efficient memory access monitoring approach for persistent memory. PMemTrace conducts best-effort pointer analysis to speculatively reduce as many redundant instrumentations to volatile writes as possible at compile time. Besides, PMemTrace enforces thread-local memory access permission control with the Memory Protection Keys (MPK) hardware primitive in recent Intel processors to track the mistakenly filtered persistent writes. The evaluation results show that PMemTrace substantially outperforms the state-of-the-art memory tracking systems. © 2023, Springer Nature Switzerland AG.

关键词： Dynamic random access storage

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：