Recommender systems are playing a significant role in modern society to alleviate the information/choice overload problem, since Internet users may feel hard to identify the most favorite items or products from millio...
详细信息
In the era of the Internet of Things(IoT),the proliferation of connected devices has raised security concerns,increasing the risk of intrusions into diverse *** the convenience and efficiency offered by IoT technology...
详细信息
In the era of the Internet of Things(IoT),the proliferation of connected devices has raised security concerns,increasing the risk of intrusions into diverse *** the convenience and efficiency offered by IoT technology,the growing number of IoT devices escalates the likelihood of attacks,emphasizing the need for robust security tools to automatically detect and explain *** paper introduces a deep learning methodology for detecting and classifying distributed denial of service(DDoS)attacks,addressing a significant security concern within IoT *** effective procedure of deep transfer learning is applied to utilize deep learning backbones,which is then evaluated on two benchmarking datasets of DDoS attacks in terms of accuracy and time *** leveraging several deep architectures,the study conducts thorough binary and multiclass experiments,each varying in the complexity of classifying attack types and demonstrating real-world ***,this study employs an explainable artificial intelligence(XAI)AI technique to elucidate the contribution of extracted features in the process of attack *** experimental results demonstrate the effectiveness of the proposed method,achieving a recall of 99.39%by the XAI bidirectional long short-term memory(XAI-BiLSTM)model.
Machine Reading Comprehension (MRC) has garnered significant attention in the field of natural language processing. While the maritime domain lacks openly available MRC task models and corresponding datasets, resultin...
详细信息
Blockchain technology has been extensively studied over the past decade as a foundation for decentralized information-sharing platforms due to its promising *** the success of existing blockchain architectures like Bi...
详细信息
Blockchain technology has been extensively studied over the past decade as a foundation for decentralized information-sharing platforms due to its promising *** the success of existing blockchain architectures like Bitcoin,Ethereum,Filecoin,Hyperledger Fabric,BCOS,and BCS,current blockchain applications are still quite *** struggles with scenarios requiring high-speed transactions(e.g.,online markets)or large data storage(e.g.,video services)due to consensus efficiency *** restrictions pose risks to investors in blockchain-based economic systems(e.g.,DeFi),deterring current and potential *** protection challenges make it difficult to involve sensitive data in blockchain applications.
Great progress has been made toward accurate face detection in recent ***,the heavy model and expensive computation costs make it difficult to deploy many detectors on mobile and embedded devices where model size and ...
详细信息
Great progress has been made toward accurate face detection in recent ***,the heavy model and expensive computation costs make it difficult to deploy many detectors on mobile and embedded devices where model size and latency are highly *** this paper,we present a millisecond-level anchor-free face detector,YuNet,which is specifically designed for edge *** are several key contributions in improving the efficiency-accuracy ***,we analyse the influential state-of-theart face detectors in recent years and summarize the rules to reduce the size of ***,a lightweight face detector,YuNet,is *** detector contains a tiny and efficient feature extraction backbone and a simplified pyramid feature fusion *** the best of our knowledge,YuNet has the best trade-off between accuracy and *** has only 75856 parameters and is less than 1/5 of other small-size *** addition,a training strategy is presented for the tiny face detector,and it can effectively train models with the same distribution of the training *** proposed YuNet achieves 81.1%mAP(single-scale)on the WIDER FACE validation hard track with a high inference efficiency(Intel i7-12700K:1.6ms per frame at 320×320).Because of its unique advantages,the repository for YuNet and its predecessors has been popular at GitHub and gained more than 11K stars at https://***/ShiqiYu/***:Face detection,object detection,computer version,lightweight,inference efficiency,anchor-free mechanism.
The Gannet Optimization Algorithm (GOA) and the Whale Optimization Algorithm (WOA) demonstrate strong performance;however, there remains room for improvement in convergence and practical applications. This study intro...
详细信息
The Gannet Optimization Algorithm (GOA) and the Whale Optimization Algorithm (WOA) demonstrate strong performance;however, there remains room for improvement in convergence and practical applications. This study introduces a hybrid optimization algorithm, named the adaptive inertia weight whale optimization algorithm and gannet optimization algorithm (AIWGOA), which addresses challenges in enhancing handwritten documents. The hybrid strategy integrates the strengths of both algorithms, significantly enhancing their capabilities, whereas the adaptive parameter strategy mitigates the need for manual parameter setting. By amalgamating the hybrid strategy and parameter-adaptive approach, the Gannet Optimization Algorithm was refined to yield the AIWGOA. Through a performance analysis of the CEC2013 benchmark, the AIWGOA demonstrates notable advantages across various metrics. Subsequently, an evaluation index was employed to assess the enhanced handwritten documents and images, affirming the superior practical application of the AIWGOA compared with other algorithms.
Feature selection (FS) is a critical step used to identify the most relevant and informative features from a given dataset. Feature selection plays a crucial role in dimensionality reduction, improving the used model...
详细信息
The approach of dividing text into chunks and building indexes is a mainstream approach for Retrieval-Augmented Generation (RAG) of Large Language Models (LLMs). However, retrieved text chunks often contain noise or r...
详细信息
ISBN:
(纸本)9798331540012
The approach of dividing text into chunks and building indexes is a mainstream approach for Retrieval-Augmented Generation (RAG) of Large Language Models (LLMs). However, retrieved text chunks often contain noise or redundant information, which can negatively impact RAG performance. To address this limitation, researchers have proposed RAG approaches based on text chunks, knowledge graphs, and long-texts. Nevertheless, there are still challenges that need to be addressed, such as the underutilization of knowledge within text chunks, the resource-intensive nature and complex process of constructing Knowledge Graphs (KG), and the lack of emphasis on optimization during the post-processing filtering stage. We propose a RAG framework with text-chunk knowledge graph (TKG-RAG). The proposed framework can construct a text-chunk knowledge graph automatically by extracting the hierarchical structure, contextual relationships, topic sentences, and inter-chunk relationships from the domain text. The framework begins by using text indexing to retrieve relevant text chunks based on similarity. These chunks are then mapped to nodes within the text-chunk knowledge graph, where connections are established based on the relationships of nodes in the graph to generate subgraphs. The content of the nodes is rearranged and merged using the hierarchical structure and relational knowledge in the graph, and then the residual isolated nodes that are not merged are fused based on the attribute knowledge in the TKG. In addition, to improve the performance of filters in the post-processing stage, we incorporate the datasets considering texts and numerical characteristics, into the fine-tuning process of the filter model. This approach aimed to enhance the accuracy of filtering and reduce token consumption in the generation stage. To validate the effectiveness of this approach, comparative and ablation experiments were conducted on five datasets: NQ, PopQA, HotpotQA, TriviaQA, and LawQA. The re
Heritage incorporates images, customs, rituals, beliefs, cultures, knowledge, arts, crafts, music, and artifacts of a region. Apart from conferring the details about the understanding of previous generations, the heri...
详细信息
Heritage incorporates images, customs, rituals, beliefs, cultures, knowledge, arts, crafts, music, and artifacts of a region. Apart from conferring the details about the understanding of previous generations, the heritage multimedia also provides details regarding innovative attitudes, ways of living, and heterogeneity of historical and archaeological approaches of a society. Since the availability of the internet in the present scenario makes an illicit user access the data easily, therefore, it is mandatory to protect the Cultural Heritage (CH) media. This paper offers a secure and resilient Discrete Cosine Transform (DCT) based blind watermarking algorithm for copyright security of digital images. We have developed and tested two different approaches of embedding namely Fixed Threshold Reference-based Relative Modulation (FTRRM) and Adaptive Threshold Reference-based Relative Modulation (ATRRM). In either approach, the watermark is embedded in the ‘Y’ component of the YCbCr color model. In FTRRM, the ‘Y’ component of the host image is first divided into 8 × 8 non-overlapping blocks and DCT is applied on each block. The watermark bits are embedded into the transformed coefficients by modulating the relative difference of coefficients depending on the bit to be embedded. A similar embedding strategy is followed in ATRRM, with a change that relative difference is modulated adaptively. Both schemes have been tested on a set of heritage images, besides general test images. We make use of chaotic and Deoxyribo Nucleic acid (DNA) encryption to ensure double-layer security of embedded watermark. The Peak Signal-to-Noise Ratio (PSNR) of the proposed scheme in the case of FTRRM is 40.4984 dB and for ATRRM it is 39.9549 dB, the Structural Similarity Index Matrix (SSIM) values are close to unity, in both cases for various test images that guarantee the imperceptibility of the proposed scheme. The robustness of each scheme is revealed by comparing them with the various state-
Various few-shot image classification methods indicate that transferring knowledge from other sources can improve the accuracy of the ***,most of these methods work with one single source or use only closely correlate...
详细信息
Various few-shot image classification methods indicate that transferring knowledge from other sources can improve the accuracy of the ***,most of these methods work with one single source or use only closely correlated knowledge *** this paper,we propose a novel weakly correlated knowledge integration(WCKI)framework to address these *** specifically,we propose a unified knowledge graph(UKG)to integrate knowledge transferred from different sources(i.e.,visual domain and textual domain).Moreover,a graph attention module is proposed to sample the subgraph from the UKG with low *** avoid explicitly aligning the visual features to the potentially biased and weakly correlated knowledge space,we sample a task-specific subgraph from UKG and append it as latent *** framework demonstrates significant improvements on multiple few-shot image classification datasets.
暂无评论