With the widespread use of UAVs in commercial and industrial applications, UAV detection is receiving increasing attention areas such as public safety. As a result, object techniques for are also developing rapidly. However, small size drones, complex airspace backgrounds, changing light conditions still pose significant challenges research this area. Based on above problems, paper proposes tiny method based optimized YOLOv8. First, head component, high-resolution added to improve device’s...

Previous chapter Next Full AccessProceedings Proceedings of the 2008 SIAM International Conference on Data Mining (SDM)Weighted Consensus ClusteringTao Li and Chris DingTao Dingpp.798 - 809Chapter DOI:https://doi.org/10.1137/1.9781611972788.72PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract clustering has emerged as an important extension classical problem. We propose weighted consensus clustering, where each input is weights are determined in...

Word embeddings carry stereotypical connotations from the text they are trained on, which can lead to invalid inferences in downstream models that rely on them. We use this observation design a mechanism for measuring stereotypes using task of natural language inference. demonstrate reduction via bias mitigation strategies static word (GloVe). Further, we show gender bias, these techniques extend contextualized when applied selectively only components (ELMo, BERT).

Clustering is the problem of identifying distribution patterns and intrinsic correlations in large data sets by partitioning points into similarity classes. This paper studies clustering binary data. case for market basket datasets where transactions contain items document documents "bag words". The contribution three-fold. First a general model presented. treats features equally, based on their symmetric association relations, explicitly describes assignments as well feature assignments. We...

Phase Change Memory (PCM) is one of the most promising technologies among emerging non-volatile memories. PCM stores data in crystalline and amorphous phases GST material using large differences their electrical resistivity. Although it possible to design a high capacity memory system by storing multiple bits at intermediate levels between highest lowest resistance states PCM, difficult obtain tight distribution required for accurate reading data. Moreover, programming latency energy...

Document clustering has long been an important problem in information retrieval. In this paper, we present a new algorithm ASI1 , which uses explicitly modeling of the subspace structure associated with each cluster. ASI simultaneously performs data reduction and identification via iterative alternating optimization procedure. Motivated from procedure, then provide novel method to determine number clusters. We also discuss connections various existential approaches. Finally, extensive...

Semi-supervised clustering (i.e., with knowledge-based constraints) has emerged as an important variant of the traditional paradigms. However, most existing semi-supervised algorithms are designed for partitional methods and few research efforts have been reported on hierarchical methods. In addition, current focused use background information in form instance level must-link cannot-link constraints, which not suitable where data objects linked over different hierarchy levels. this paper, we...

Clustering is an old research topic in data mining and machine learning. Most of the traditional clustering methods can be categorized as local or global ones. In this paper, a novel method that explore both information set proposed. The method, with Local Global Regularization (CLGR), aims to minimize cost function properly trades off costs. We show such optimization problem solved by eigenvalue decomposition sparse symmetric matrix, which done efficiently using iterative methods. Finally,...

Ensemble clustering has emerged as an important elaboration of the classical problems. refers to situation in which a number different (input) clusterings have been obtained for particular dataset and it is desired find single (consensus) better fit some sense than existing clusterings. Many approaches developed solve ensemble problems over last few years. However, most these techniques are designed partitional methods. Few research efforts reported hierarchical In this paper, we propose...

In recent years, document clustering has been receiving more and attentions as an important fundamental technique for unsupervised organization, automatictopic extraction, fast information retrieval or filtering. this paper, we propose a novel method documents using regularization. Unlike traditional globally regularized methods, our first construct local linear label predictor each vector, then combine all those regularizers with global smoothness regularizer. So call algorithm Clustering...

Previous chapter Next Full AccessProceedings Proceedings of the 2009 SIAM International Conference on Data Mining (SDM)Integrated KL (K-means – Laplacian) Clustering: A New Clustering Approach by Combining Attribute and Pairwise RelationsFei Wang, Chris Ding, Tao LiFei Lipp.38 - 48Chapter DOI:https://doi.org/10.1137/1.9781611972795.4PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract Most datasets in real applications come from multiple sources....

Ensemble clustering, as an important extension of the clustering problem, refers to problem combining different (input) clusterings a given dataset generate final (consensus) that is better fit in some sense than existing clusterings. Over past few years, many ensemble approaches have been developed. However, most them are designed for partitional methods, and research efforts reported hierarchical methods. In this article, framework can naturally combine both results proposed. addition,...

Convolutional neural networks (CNNs) have been successfully applied in artificial intelligent systems to perform sensory processing, sequence learning, and image processing. In contrast conventional computing-centric applications, CNNs are known be both computationally memory intensive. The computational resources of CNN applications mixed together the network weights. This incurs a significant amount data movement, especially for high-dimensional convolutions. emerging Processing-in-Memory...

Deep convolutional neural networks (CNNs) are widely adopted in intelligent systems with unprecedented accuracy but at the cost of a substantial amount data movement. Although emerging processing-in-memory (PIM) architecture seeks to minimize movement by placing memory near processing elements, is still major bottleneck entire system. The selection hyper-parameters training CNN applications requires over hundreds kilobytes cache capacity for concurrent convolutions. How jointly explore...

The emerging of deep neural networks, especially the convolutional network (CNN), substantially promotes fast development brainware processors in object detection. However, vast architecture brings severe challenges to design processor, which requires a large number logic gates and memories. Therefore, compact processor with less memory gate has high demand Typically, detection involves single-shot multi-shot detectors accordance different principle. In early stage, detector leading role...