large scale retrieval Research Articles (Page 1)

With the increasing number of network security threats and the frequent occurrence of software vulnerability attacks, the effective management and large-scale retrieval of vulnerability data have become urgent needs. Existing vulnerability information is scattered across heterogeneous sources and is difficult to integrate, which in turn makes it hard for security analysts to quickly retrieve and analyze relevant security knowledge. To address this problem, this paper proposes a method to construct a vulnerability knowledge graph by integrating multi-source vulnerability data, combining graph embedding technology with large language model reasoning to aggregate, infer, and enrich vulnerability knowledge. Experiments demonstrated that our domain-tuned Bidirectional Long Short-Term Memory–Conditional Random Field (BiLSTM-CRF) named entity recognition (NER), enhanced with a cybersecurity dictionary, achieved a 90.1% F1-score for entity extraction. For link prediction, a hybrid Graph Attention Network fused with GPT-3 reasoning boosted Hits1 by 0.137, Hits3 by 0.116, and Hits10 by 0.101 over the baseline. These results confirm that our approach markedly enhanced entity identification and relationship inference, yielding a more complete and dynamically updatable cybersecurity knowledge graph.

In response to the shortcomings of high computational cost and low accuracy of the local recognition descriptor method for palmprint images, this paper proposes a new Local Ordinal Code (LOC), which utilizes three common filters for multi-directional filtering coding. For the demand for large-scale retrieval, a new dimension control factor is proposed to perform linear feature dimensionality reduction to achieve fast matching. In addition, Feature Fusion coding schemes of LOC (FFLOC) are proposed on the basis of LOC, and the obtained feature codes are fused with different weights to achieve higher recognition accuracy. The algorithm proposed in this paper is tested in five publicly available palmprint databases and one self-collected palmprint database, and the best results are obtained on most datasets.

The traditional model upgrading paradigm for retrieval requires recomputing all gallery embeddings before deploying the new model (dubbed as "backfilling"), which is quite expensive and time-consuming considering billions of instances in industrial applications. BCT presents the first step towards backward-compatible model upgrades to get rid of backfilling. It is workable but leaves the new model in a dilemma between new feature discriminativeness and new-to-old compatibility due to the undifferentiated compatibility constraints. In this work, we propose Darwinian Model Upgrades (DMU), which disentangle the inheritance and variation in the model evolving with selective backward compatibility and forward adaptation, respectively. The old-to-new heritable knowledge is measured by old feature discriminativeness, and the gallery features, especially those of poor quality, are evolved in a lightweight manner to become more adaptive in the new latent space. We demonstrate the superiority of DMU through comprehensive experiments on large-scale landmark retrieval and face recognition benchmarks. DMU effectively alleviates the new-to-new degradation at the same time improving new-to-old compatibility, rendering a more proper model upgrading paradigm in large-scale retrieval systems.Code: https://github.com/TencentARC/OpenCompatible.

The high efficiency in computation and storage makes hashing (including binary hashing and quantization) a common strategy in large-scale retrieval systems. To alleviate the reliance on expensive annotations, unsupervised deep hashing becomes an important research problem. This paper provides a novel solution to unsupervised deep quantization, namely Contrastive Quantization with Code Memory (MeCoQ). Different from existing reconstruction-based strategies, we learn unsupervised binary descriptors by contrastive learning, which can better capture discriminative visual semantics. Besides, we uncover that codeword diversity regularization is critical to prevent contrastive learning-based quantization from model degeneration. Moreover, we introduce a novel quantization code memory module that boosts contrastive learning with lower feature drift than conventional feature memories. Extensive experiments on benchmark datasets show that MeCoQ outperforms state-of-the-art methods. Code and configurations are publicly released.

Fast and Accurate Normal Estimation for Point Clouds Via Patch Stitching

Owing to the significant societal value of inland water resources, there is a need for cost-effective monitoring of water quality on large scales. We tested the suitability of the recently launched Sentinel-2A to monitor a key water quality parameter, coloured dissolved organic matter (CDOM), in various types of lakes in northern Sweden. Values of a(420)CDOM (CDOM absorption at 420 nm wavelength) were obtained by analyzing water samples from 46 lakes in five districts across Sweden within an area of approximately 800 km2. We evaluated the relationships between a(420)CDOM and band ratios derived from Sentinel-2A Level-1C and Level-2A products. The band ratios B2/B3 (460 nm/560 nm) and B3/B5 (560 nm/705 nm) showed poor relationships with a(420)CDOM in Level-1C and 2A data both before and after the removal of outliers. However, there was a slightly stronger power relationship between the atmospherically-corrected B3/B4 ratio and a(420)CDOM (R2 = 0.28, n = 46), and this relationship was further improved (R2 = 0.65, n = 41) by removing observations affected by light haze and cirrus clouds. This study covered a wide range of lakes in different landscape settings and demonstrates the broad applicability of a(420)CDOM retrieval algorithms based on the B3/B4 ratio derived from Sentinel-2A.

Motion capture data describe human movements in the form of spatio-temporal trajectories of skeleton joints. Intelligent management of such complex data is a challenging task for computers which requires an effective concept of motion similarity. However, evaluating the pair-wise similarity is a difficult problem as a single action can be performed by various actors in different ways, speeds or starting positions. Recent methods usually model the motion similarity by comparing customized features using distance-based functions or specialized machine-learning classifiers. By combining both these approaches, we transform the problem of comparing motions of variable sizes into the problem of comparing fixed-size vectors. Specifically, each rather-short motion is encoded into a compact visual representation from which a highly descriptive 4,096-dimensional feature vector is extracted using a fine-tuned deep convolutional neural network. The advantage is that the fixed-size features are compared by the Euclidean distance which enables efficient motion indexing by any metric-based index structure. Another advantage of the proposed approach is its tolerance towards an imprecise action segmentation, the variance in movement speed, and a lower data quality. All these properties together bring new possibilities for effective and efficient large-scale retrieval.

With the advent of commodity 3D capturing devices and better 3D modeling tools, 3D shape content is becoming increasingly prevalent. Therefore, the need for shape retrieval algorithms to handle large-scale shape repositories is more and more important. This track provides a benchmark to evaluate large-scale 3D shape retrieval based on the ShapeNet dataset. It is a continuation of the SHREC 2016 large-scale shape retrieval challenge with a goal of measuring progress with recent developments in deep learning methods for shape retrieval. We use ShapeNet Core55, which provides more than 50 thousands models over 55 common categories in total for training and evaluating several algorithms. Eight participating teams have submitted a variety of retrieval methods which were evaluated on several standard information retrieval performance metrics. The approaches vary in terms of the 3D representation, using multi-view projections, point sets, volumetric grids, or traditional 3D shape descriptors. Overall performance on the shape retrieval task has improved significantly compared to the iteration of this competition in SHREC 2016. We release all data, results, and evaluation code for the benefit of the community and to catalyze future research into large-scale 3D shape retrieval (website: https://www.shapenet.org/shrec17).

What is the story of an image? What is the relationship between pictures, language, and information we can extract using state of the art computational recognition systems? In an attempt to address both of these questions, we explore methods for retrieving and generating natural language descriptions for images. Ideally, we would like our generated textual descriptions (captions) to both sound like a person wrote them, and also remain true to the image content. To do this we develop data-driven approaches for image description generation, using retrieval-based techniques to gather either: (a) whole captions associated with a visually similar image, or (b) relevant bits of text (phrases) from a large collection of image + description pairs. In the case of (b), we develop optimization algorithms to merge the retrieved phrases into valid natural language sentences. The end result is two simple, but effective, methods for harnessing the power of big data to produce image captions that are altogether more general, relevant, and human-like than previous attempts.

The traditional thesaurus have rich vocabulary and clear semantic structure which is easy to be extracted concepts and relationships, providing great convenience for ontology building. Chinese thesaurus is a large-scale comprehensive science and technology retrieval tools which has its own characteristics and the vocabulary range includes all fields such as natural science, medicine, agriculture and engineering technology. Domain ontology construction from Chinese Thesaurus need to clear the definition and semantic relation of descriptors, including two important tasks: the core concept selection and the transformation of relationship between concepts.

The major public microarray repositories Gene Expression Omnibus and ArrayExpress are growing rapidly. This enables meta-analysis studies, in which expression data from multiple individual studies are combined. To facilitate these types of studies, we developed Microarray Retriever for searching and retrieval of data from GEO and ArrayExpress. The tool allows access to the two repositories simultaneously, to search in the repositories using complex queries, to retrieve microarray data for published articles and to download data in one structured archive. The tool is available on the web at: http://www.lgtc.nl/MaRe/

Abstract Author cocitations form the base of a general technique of bibliogrpahic retrieval and of mapping the strucutre of literatures. The technique was used here in a largescale retrieval of the Social Indicators (SI) literature and a computerized mapping, through multidimensional scaling and clustering, of key SI contributors as perceived by citers during 1972‐1980. Clusters of points on the map are interpreted as authors similar in perceived subject matter and style of work; other interpretable features also eemerge. The map conforms well with independent reviews of the SI literature, actually illustrating some of the judgments they contain. It is suggested that such mappings may become a stimulus and complement to literature reviews in the future.

Abstract This paper reviews the general problem of finding an equation of state for a pure substance, and discusses the various ways of analyzing the isochores. The approach used was to replace the conventional isometric diagram showing the vapour pressure curve with either a ZT, versus Tr isometric diagram or a P r/Z versus Pr isometric diagram. The latter diagrams show the two-phase envelope which permits use of the saturated state for the origin of individual isochors. This results in the density implicit formulations. (Equation Available In Full Paper) which may be conveniently solved with a digital computer. To demonstrate the method, analytical expressions were developed to represent the compressibility factor tables of Lydersen et a1. For reduced densities greater than one, in the reduced temperature range from 0.3 to 3.0, and the pressure range from saturation to Pr = 30, a standard relative error of 0.65 per cent was obtained with only five constants. The approach is also applicable for reduced densities of less than one, but, at the present time, owing to an inability to 6accurately describe the saturated vapour curve in terms of density, the results are only qualitatively Correct. The sparsity of data for Ze values of 0.25 and 0.29 in this region prevented development of a cross-correlation with Z. INTRODUCTION COMPUTER PROGRAMS used for modelling the performance of petroleum reservoirs, or for describing chemical other industrial processes, frequently require a method for calculating fluid densities. This is usually accomplished through the use of a suitable equation of state or through corresponding states correlations, The latter are often used to estimate the density of pure substances, and they can be extended to mixture through the use of the concept of pseudocritical properties. In order to use corresponding states correlations, a suitable method must be found to adapt the information to computer use. This may be accomplished by applying interpolation techniques to tabular data which have been read into memory, by fitting empirical functions to all or part of the data or by using the reduced form of an equation of state. Of these three possibilities, the use of interpolation techniques is considered to be the least desirable, because large-scale data storage and subsequent retrieval is an inefficient computer operation. The well-known equations of Beattie and Bridgeman(14), and Benedict, Webb and Rubin(15) have been applied in reduced form to generalized correlations by Su and Viswanath(18) Good accuracy has been achieved using these equations in the gaseous region, but unfortunately it has not been possible to find a single set of coefficients that will give satisfactory results over the entire liquid, gaseous and saturated regions. Hirschfolder, Buehler, McGee and Sutton(7) have developed a generalized equation of state in which different functions are used for the three regions defined as:gas, where the reduced density is less than one;