Streaming media is pervasive on the Internet now and continues to grow rapidly. A large scale streaming media system may demand thousands of disks to satisfy both the bandwidth and storage capacity requirements impose...
详细信息
Streaming media is pervasive on the Internet now and continues to grow rapidly. A large scale streaming media system may demand thousands of disks to satisfy both the bandwidth and storage capacity requirements imposed by thousands of concurrent clients. While performance has long been the focus of large scale storage systems research, manageability has become the dominant criterion in evaluating storage solutions, as the cost of storage management outweighs the cost of the storage devices themselves by a factor of three to eight. This paper designs and implements a highly reliable and available distributed media data oriented innovative Storage Management Mechanism (SMM) in a scalable streaming media system. All heterogeneous storage resources across the system are aggregated as a single logical view which strikes a good balance between a distributed data storage and centralized storage management. The centralized SMM dramatically simplifies the storage administration as the size and complexity of the streaming media system grows. Fault tolerance and dynamic load balance at the disk level and system level are achieved to provide high reliability and availability due to the SMM.
With a SAN (Storage Area Network), large-capacity storage can be shared among multiple hosts at high speed. The potential market for SAN is enormous, but the technology won't become ubiquitous overnight. The compl...
详细信息
With a SAN (Storage Area Network), large-capacity storage can be shared among multiple hosts at high speed. The potential market for SAN is enormous, but the technology won't become ubiquitous overnight. The complex administration is indicated by many researches as a key barrier to adopt SAN solutions. This paper designs and implements a virtual storage image (VSI) that aggregates storage space located on heterogeneous storage nodes (see Fig.1)in a Fibre Channel point-to-point (FC-P2P) SAN.
Federated Learning (FL) has emerged as a promising paradigm for training machine learning models across distributed devices while preserving their data privacy. However, the robustness of FL models against adversarial...
详细信息
Federated Learning (FL) has emerged as a promising paradigm for training machine learning models across distributed devices while preserving their data privacy. However, the robustness of FL models against adversarial data and model attacks, noisy updates, and label-flipped data issues remain a critical concern. In this paper, we present a systematic literature review using the PRISMA framework to comprehensively analyze existing research on robust FL. Through a rigorous selection process using six key databases (ACM Digital Library, IEEE Xplore, ScienceDirect, Springer, Web of Science, and Scopus), we identify and categorize 244 studies into eight themes of ensuring robustness in FL: objective regularization, optimizer modification, differential privacy employment, additional dataset requirement and decentralization orchestration, manifold, client selection, new aggregation algorithms, and aggregation hyperparameter tuning. We synthesize the findings from these themes, highlighting the various approaches and their potential gaps proposed to enhance the robustness of FL models. Furthermore, we discuss future research directions, focusing on the potential of hybrid approaches, ensemble techniques, and adaptive mechanisms for addressing the challenges associated with robust FL. This review not only provides a comprehensive overview of the state-of-the-art in robust FL but also serves as a roadmap for researchers and practitioners seeking to advance the field and develop more robust and resilient FL systems.
This book constitutes the refereed proceedings of six workshops of the 14th International Conference on Web-Age Information Management, WAIM 2013, held in Beidaihe, China, June 2013. The 37 revised full papers are org...
详细信息
ISBN:
(数字)9783642395277
ISBN:
(纸本)9783642395260
This book constitutes the refereed proceedings of six workshops of the 14th International Conference on Web-Age Information Management, WAIM 2013, held in Beidaihe, China, June 2013. The 37 revised full papers are organized in topical sections on the six following workshops: The International Workshop on Big Data Management on Emerging Hardware (HardBD 2013), the Second International Workshop on Massive Data Storage and Processing (MDSP 2013), the First International Workshop on Emergency Management in Big Data Age (BigEM 2013), the International Workshop on Trajectory Mining in Social Networks (TMSN 2013), the First International Workshop on Location-based Query Processing in Mobile Environments (LQPM 2013), and the First International Workshop on Big Data Management and service (BDMS 2013).
Non-overlapping Cross-domain Sequential Recommendation (NCSR) is the task that focuses on domain knowledge transfer without overlapping entities. Compared with traditional Cross-domain Sequential Recommendation (CSR),...
详细信息
Non-overlapping Cross-domain Sequential Recommendation (NCSR) is the task that focuses on domain knowledge transfer without overlapping entities. Compared with traditional Cross-domain Sequential Recommendation (CSR), NCSR poses several challenges: 1) NCSR methods often rely on explicit item IDs, overlooking semantic information among entities. 2) Existing CSR mainly relies on domain alignment for knowledge transfer, risking semantic loss during alignment. 3) Most previous studies do not consider the many-to-one characteristic, which is challenging because of the utilization of multiple source domains. Given the above challenges, we introduce the prompt learning technique for Many-to-one Non-overlapping Cross-domain Sequential Recommendation (MNCSR) and propose a Text-enhanced Co-attention Prompt Learning Paradigm (TCPLP). Specifically, we capture semantic meanings by representing items through text rather than IDs, leveraging natural language universality to facilitate cross-domain knowledge transfer. Unlike prior works that need to conduct domain alignment, we directly learn transferable domain information, where two types of prompts, i.e., domain-shared and domain-specific prompts, are devised, with a co-attention-based network for prompt encoding. Then, we develop a two-stage learning strategy, i.e., pre-train & prompt-tuning paradigm, for domain knowledge pre-learning and transferring, respectively. We conduct extensive experiments on three datasets and the experimental results demonstrate the superiority of our TCPLP. Our source codes have been publicly released.
暂无评论