Underlying Data Structure Research Articles

Weakly-supervised Semantic Guided Hashing for Social Image Retrieval

Hashing has been widely investigated for large-scale image retrieval due to its search effectiveness and computation efficiency. In this work, we propose a novel Semantic Guided Hashing method coupled with binary matrix factorization to perform more effective nearest neighbor image search by simultaneously exploring the weakly-supervised rich community-contributed information and the underlying data structures. To uncover the underlying semantic information from the weakly-supervised user-provided tags, the binary matrix factorization model is leveraged for learning the binary features of images while the problem of imperfect tags is well addressed. The uncovered semantic information enables to well guide the discrete hash code learning. The underlying data structures are discovered by adaptively learning a discriminative data graph, which makes the learned hash codes preserve the meaningful neighbors. To the best of our knowledge, the proposed method is the first work that incorporates the hash code learning, the semantic information mining and the data structure discovering into one unified framework. Besides, the proposed method is extended to one deep approach for the optimal compatibility of discriminative feature learning and hash code learning. Experiments are conducted on two widely-used social image datasets and the proposed method achieves encouraging performance compared with the state-of-the-art hashing methods.

Deep Fuzzy Hashing Network for Efficient Image Retrieval

Hashing methods for efficient image retrieval aim at learning hash functions that map similar images to semantically correlated binary codes in the Hamming space with similarity well preserved. The traditional hashing methods usually represent image content by hand-crafted features. Deep hashing methods based on deep neural network (DNN) architectures can generate more effective image features and obtain better retrieval performance. However, the underlying data structure is hardly captured by existing DNN models. Moreover, the similarity (either visually or semantically) between pairwise images is ambiguous, even uncertain, to be measured in the existing deep hashing methods. In this article, we propose a novel hashing method termed deep fuzzy hashing network (DFHN) to overcome the shortcomings of existing deep hashing approaches. Our DFHN method combines the fuzzy logic technique and the DNN to learn more effective binary codes, which can leverage fuzzy rules to model the uncertainties underlying the data. Derived from fuzzy logic theory, the generalized hamming distance is devised in the convolutional layers and fully connected layers in our DFHN to model their outputs, which come from an efficient <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">xor</small> operation on given inputs and weights. Extensive experiments show that our DFHN method obtains competitive retrieval accuracy with highly efficient training speed compared with several state-of-the-art deep hashing approaches on two large-scale image datasets: CIFAR-10 and NUS-WIDE.

Chemometrics in comprehensive two-dimensional liquid chromatography: A study of the data structure and its multilinear behavior

Comprehensive multidimensional chromatographic techniques, such as GC×GC coupled to FID or MS and LC×LC coupled to UV or MS, have gained popularity in recent years. From the analytical perspective, these techniques allow obtaining higher peak capacities and resolution power, as well as adding selectivity from the second orthogonal dimension. From the chemometric point of view, these multidimensional techniques generate highly complex datasets, which present several challenges for their analysis. On the one hand, the selection of the appropriate chemometric data analysis tool requires the understanding of the underlying data structure and its multilinear behavior. On the other hand, peak resolution in complex samples is still a challenge, because of their possible overlapping in one or two chromatographic dimensions despite the increased resolution power.In this work, a comprehensive two-dimensional liquid chromatography method hyphenated simultaneously to PDA and MS detectors was employed for the analysis of a mixture of 31 pharmaceutical compounds. Chemometric evaluation of the obtained two-dimensional chromatograms focuses on two different goals. First, the assessment of the multilinear behavior of the high-dimensional data for each of the two detection modes (LC×LC-UV and LC×LC-MS) and, also, for the multiset data obtained by fusion of the data coming from both detectors. In addition, the chemometric resolution of peaks of overlapping compounds was evaluated using the multivariate curve resolution alternating least squares (MCR-ALS) method. Finally, the advantages of data fusion from UV and MS detectors were discussed, such as the increased ability for compound identification.

An Efficient Sum Query Algorithm for Distance-Based Locally Dominating Functions

In this paper, we consider the following sum query problem: Given a point set P in $${\mathbb {R}}^d$$ , and a distance-based function f(p, q) (i.e., a function of the distance between p and q) satisfying some general properties, the goal is to develop a data structure and a query algorithm for efficiently computing a $$(1+\epsilon )$$ -approximate solution to the sum $$\sum _{p \in P} f(p,q)$$ for any query point $$q \in {\mathbb {R}}^d$$ and any small constant $$\epsilon >0$$ . Existing techniques for this problem are mainly based on some core-set techniques which often have difficulties to deal with functions with local domination property. Based on several new insights to this problem, we develop in this paper a novel technique to overcome these encountered difficulties. Our algorithm is capable of answering queries with high success probability in time no more than $${\tilde{O}}_{\epsilon ,d}(n^{0.5 + c})$$ , and the underlying data structure can be constructed in $${\tilde{O}}_{\epsilon ,d}(n^{1+c})$$ time for any $$c>0$$ , where the hidden constant has only polynomial dependence on $$1/\epsilon$$ and d. Our technique is simple and can be easily implemented for practical purpose.

BASYS-MTB: An integrative structural simulation platform based on adjacency matrices

The objective of this work is to develop and test a new data structure, which takes advantage of the latest programming technology to build interactive structural engineering modeling software that is robust, efficient, portable, modular, expandable, and easy to use. The data structure proposed is composed of a collection of matrices that describe both the geometrical properties and the topological adjacency information of the structural model, as well as the section, material, loading, and fixity attributes of the structural elements. This data structure is dubbed the Adjacency Matrix Data Structure (AMDS). There are eight matrices in the AMDS, each capturing the details of one type of structural entity or attribute: Vertex, Edge, Face, Group, Section, Material, Load, and Fixity. The design of the AMDS allows it to excel in terms of efficiency and speed when used in a matrix-based programming environment, such as MATLAB. The final outcome called “BASYS-MTB” is a structural modeling and visualization MATLAB Toolbox that is a demonstration implementation of the AMDS as the underlying data structure. The toolbox, which can serve as a preprocessor, solver, and postprocessor, employs the object-oriented programming capabilities in MATLAB to create and manipulate numerical and graphical objects. This system offers engineers and researchers with limited programming expertise a structural modeler that can be easily customized to their requirements using their existing knowledge in MATLAB.

Nearest-neighbor Projected-Distance Regression (NPDR) for detecting network interactions with adjustments for multiple tests and confounding.

Machine learning feature selection methods are needed to detect complex interaction-network effects in complicated modeling scenarios in high-dimensional data, such as GWAS, gene expression, eQTL and structural/functional neuroimage studies for case-control or continuous outcomes. In addition, many machine learning methods have limited ability to address the issues of controlling false discoveries and adjusting for covariates. To address these challenges, we develop a new feature selection technique called Nearest-neighbor Projected-Distance Regression (NPDR) that calculates the importance of each predictor using generalized linear model regression of distances between nearest-neighbor pairs projected onto the predictor dimension. NPDR captures the underlying interaction structure of data using nearest-neighbors in high dimensions, handles both dichotomous and continuous outcomes and predictor data types, statistically corrects for covariates, and permits statistical inference and penalized regression. We use realistic simulations with interactions and other effects to show that NPDR has better precision-recall than standard Relief-based feature selection and random forest importance, with the additional benefit of covariate adjustment and multiple testing correction. Using RNA-Seq data from a study of major depressive disorder (MDD), we show that NPDR with covariate adjustment removes spurious associations due to confounding. We apply NPDR to eQTL data to identify potentially interacting variants that regulate transcripts associated with MDD and demonstrate NPDR's utility for GWAS and continuous outcomes. Available at: https://insilico.github.io/npdr/. Supplementary data are available at Bioinformatics online.

Single image super-resolution via low-rank tensor representation and hierarchical dictionary learning

Super-resolution (SR) has been widely studied due to its importance in real applications and scenarios. In this paper, we focus on generating an SR image from a single low-resolution (LR) input image by employing the multi-resolution structures of an input image. By taking the LR image and its downsampled resolution (DR) and upsampled resolution (UR) versions as inputs, we propose a hierarchical dictionary learning approach to learn the latent UR-LR dictionary pair by preserving the internal structure coherence with the LR-DR dictionary pair. Note that an imposed restriction involved in this process is that the pairwise resolution images are jointly trained to obtain more compact patterns of image patches. In particular, to better explore the underlying structures of tensor data spanned by image patches, we propose a low-rank tensor approximation (LRTA) algorithm based on nuclear-norm regularization to embed input image patches into a low-dimensional space. Experimental results from publicly used images show that our proposed method achieves performance comparable with that of other state-of-the-art SR algorithms, even without using any external training databases.

Semi-Paired Asymmetric Deep Cross-Modal Hashing Learning

Hashing has been effectively applied in large-scale multimedia retrieval tasks due to the characteristics of fast calculation speed and low storage cost. However, most existing cross-view hashing methods require well paired views, where one sample in one view can always be associated with one sample in another view. Nevertheless, such fully-paired setting is difficult to hold in practice. The cross-view retrieval task for semi-paired data, in which only a portion of the sample is paired, remains more challenging and less explored. In this case, we design an unsupervised deep hashing for large-scale data, named semi-paired asymmetric deep cross-model hashing (SADCH) to address the challenging task. SADCH is a novel asymmetric end-to-end deep neural network model. Specifically, SADCH trains deep network by using query points to improve the training efficiency, and directly learns hash codes of database. A similarity matrix is constructed by a novel cross-view graph to explore the underlying data structures, such that the similarities of paired points and unpaired points can be preserved. As such, the deep features and similarity graph matrix can be jointly used to design an alternating algorithm to efficiently generate more differentiated hash codes. SADCH is evaluated for large-scale cross-view approximate nearest neighbor search on three benchmark datasets and compared against several state-of-the arts. The results of extensive experiments demonstrate the superiority of the proposed SADCH for semi-paired hashing in the unsupervised setting.

B-mode Ultrasound Measurement of Facial Soft Tissue Thickness for Craniofacial Identification: A Standardized Approach.

Facial soft tissue thicknesses (FSTT) have long formed a quantitative cornerstone of craniofacial identification methods. Measurement approaches could, however, be improved by standardization that enhances the utility/comparability of these FSTT data between studies and authors. This applies equally within the broad classes of measurement techniques, just as it does between them, where many different varieties of tools, tool settings, and practitioner techniques are used within each measurement modality. Although B-mode ultrasound is popularly used and holds some prime advantages, such as the measurement of upright living subjects, technical recommendations that provide basic underlying data structure and standardization are essentially nonexistent. This paper provides the first systematic and illustrated description of a standardized B-mode ultrasound measurement method designed to maximize data utility for craniofacial identification purposes.

Graph Structure Fusion for Multiview Clustering

Most existing multiview clustering methods take graphs, which are usually predefined independently in each view, as input to uncover data distribution. These methods ignore the correlation of graph structure among multiple views and clustering results highly depend on the quality of predefined affinity graphs. We address the problem of multiview clustering by seamlessly integrating graph structures of different views to fully exploit the geometric property of underlying data structure. The proposed method is based on the assumption that the intrinsic underlying graph structure would assign corresponding connected component in each graph to the same cluster. Different graphs from multiple views are integrated by using the Hadamard product since different views usually together admit the same underlying structure across multiple views. Specifically, these graphs are integrated into a global one and the structure of the global graph is adaptively tuned by a well-designed objective function so that the number of components of the graph is exactly equal to the number of clusters. It is worth noting that we directly obtain cluster indicators from the graph itself without performing further graph-cut or $k$k-means clustering algorithms. Experiments show the proposed method obtains better clustering performance than the state-of-the-art methods.

Cost-sensitive dual-bidirectional linear discriminant analysis

In most previous cost-sensitive feature extraction methods, the image matrix needs to be converted into vectors. The conversion always leads to a high computation complexity and small sample size problem. To address these issues, we propose a matrix-feature extraction method for face recognition, Cost-sensitive Dual-Bidirectional Linear Discriminant Analysis (CB2LDA). It is based on 2D image matrices, which greatly reduces the computation complexity and the probability of falling into small sample size problems. The proposed methods extract 2D matrix features from a diagonal block matrix containing both image matrix A and its transposition AT. With the block matrix, the scatter information in both directions is simultaneously considered in the projections, which helps to preserve the underlying data structure in images. Moreover, it aims to preserve the best cost-weighted discriminative information in the facial images, such that the misclassification costs reach a lower level. The experimental results validate the effectiveness and efficiency of the proposed method.

Concurrent Treaps and Impact of Locking Objects

We propose algorithms to perform operations concurrently on treaps in a shared memory multi-core environment. Concurrent treaps hold the advantage of using nodes’ priority for maintaining the height of the treaps. To achieve synchronization, concurrent treaps make use of fine-grained locking mechanism along with logical ordering and physical ordering of nodes’ keys. We initially study the throughput and performance-per-Watt (PPW) of our concurrent treap implementation and observe that it scales well, and performs better than the state-of-the-art implementations. We further continue studies to understand the impact of different locking objects on both throughput and PPW. Our experiments show that a concurrent treap implementation that uses AtomicInteger as the locking object provides better throughput and PPW, at the same time uses a low memory footprint. As part of the application study, we consider concurrent interval trees by choosing different underlying concurrent search tree implementations as the base. We observe that the concurrent interval tree implementation that uses concurrent treap as an underlying data structure provides better throughput.

A data structure perspective to the RDD-based Apriori algorithm on Spark

During the recent years, a number of efficient and scalable frequent itemset mining algorithms for big data analytics have been proposed by many researchers. Initially, MapReduce-based frequent itemset mining algorithms on Hadoop cluster were proposed. Although, Hadoop has been developed as a cluster computing system for handling and processing big data, but the performance of Hadoop does not meet the expectation for the iterative algorithms of data mining, due to its high I/O, and writing and then reading intermediate results in the disk. Consequently, Spark has been developed as another cluster computing infrastructure which is much faster than Hadoop due to its in-memory computation. It is highly suitable for iterative algorithms and supports batch, interactive, iterative, and stream processing of data. Many frequent itemset mining algorithms have been re-designed on the Spark, and most of them are Apriori-based. All these Spark-based Apriori algorithms use Hash Tree as the underlying data structure. This paper investigates the efficiency of various data structures for the Spark-based Apriori. Although, the data structure perspective has been investigated previously, but for MapReduce-based Apriori, and it must be re-investigated in the distributed computing environment of Spark. The considered underlying data structures are Hash Tree, Trie, and Hash Table Trie. The experimental results on the benchmark datasets show that the performance of Spark-based Apriori with Trie and Hash Table Trie are almost similar but both perform many times better than Hash Tree in the distributed computing environment of Spark.

Visualizing probabilistic models and data with Intensive Principal Component Analysis

Unsupervised learning makes manifest the underlying structure of data without curated training and specific problem definitions. However, the inference of relationships between data points is frustrated by the "curse of dimensionality" in high dimensions. Inspired by replica theory from statistical mechanics, we consider replicas of the system to tune the dimensionality and take the limit as the number of replicas goes to zero. The result is intensive embedding, which not only is isometric (preserving local distances) but also allows global structure to be more transparently visualized. We develop the Intensive Principal Component Analysis (InPCA) and demonstrate clear improvements in visualizations of the Ising model of magnetic spins, a neural network, and the dark energy cold dark matter ([Formula: see text]) model as applied to the cosmic microwave background.

Aligned Dynamic-Preserving Embedding for Zero-Shot Action Recognition

Zero-shot learning (ZSL) typically explores a shared semantic space in order to recognize novel categories in the absence of any labeled training data. However, the traditional ZSL methods always suffer from serious domain shift problem in human action recognition. This is because: 1) existing ZSL methods are specifically designed for object recognition from static images, which do not capture the temporal dynamics of video sequences, and poor performances are always generated if those methods are directly applied to zero-shot action recognition; 2) these methods always blindly project the target data into a shared space using a semantic mapping obtained by the source data without any adaptation, in which the underlying structures of target data are ignored; and 3) severe inter-class variations exist in various action categories. The traditional ZSL methods do not take relationships across different categories into consideration. In this paper, we propose a novel aligned dynamic-preserving embedding (ADPE) model for zero-shot action recognition in a transductive setting. In our model, an adaptive embedding of target videos is learned, exploring the distributions of both the source and target data. An aligned regularization is further proposed to couple the centers of target semantic representations with their corresponding label prototypes in order to preserve the relationships across different categories. Most significantly, during our embedding, the temporal dynamics of video sequences are simultaneously preserved via exploiting the temporal consistency of video sequences and capturing the temporal evolution of successive segments of actions. Our model can effectively overcome the domain shift problem in zero-shot action recognition. The experiments on Olympic sports, HMDB51, and UCF101 datasets demonstrate the effectiveness of our model.

Context Definition and Query Language: Conceptual Specification, Implementation, and Evaluation.

As IoT grows at a staggering pace, the need for contextual intelligence is a fundamental and critical factor for IoT intelligence, efficiency, effectiveness, performance, and sustainability. As the standardisation efforts for IoT are fast progressing, efforts in standardising context management platforms led by the European Telecommunications Standards Institute (ETSI) are gaining more attention from both academic and industrial research organizations. These standardisation endeavours will enable intelligent interactions between ‘things’, where things could be devices, software components, web-services, or sensing/actuating systems. Therefore, having a generic platform to describe and query context is crucial for the future of IoT applications. In this paper, we propose Context Definition and Query Language (CDQL), an advanced approach that enables things to exchange, reuse and share context between each other. CDQL consists of two main parts, namely: context definition model, which is designed to describe situations and high-level context; and Context Query Language (CQL), which is a powerful and flexible query language to express contextual information requirements without considering details of the underlying data structures. An important feature of the proposed query language is its ability to query entities in IoT environments based on their situation in a fully dynamic manner where users can define situations and context entities as part of the query. We exemplify the usage of CDQL on three different smart city use cases to highlight how CDQL can be utilised to deliver contextual information to IoT applications. Performance evaluation has demonstrated scalability and efficiency of CDQL in handling a fairly large number of concurrent context queries.

Restricted-derestricted dynamic Bayesian Network inference of transcriptional regulatory relationships among genes in cancer

Understanding transcriptional regulatory relationships among genes is important for gaining etiological insights into diseases such as cancer. To this end, high-throughput biological data have been generated through advancements in a variety of technologies. These rely on computational approaches to discover underlying structures in such data. Among these computational approaches, Bayesian networks (BNs) stand out because their probabilistic nature enables them to manage randomness in the dynamics of gene regulation and experimental data. Feedback loops inherent in networks of regulatory relationships are more tractable when enhancements to BNs are applied to them. Here, we propose Restricted-Derestricted dynamic BNs with a novel search technique, Restricted-Derestricted Greedy Method, for such tasks. This approach relies on the Restricted-Derestricted Greedy search technique to infer transcriptional regulatory networks in two phases: restricted inference and derestricted inference. An application of this approach to real data sets reveals it performs favourably well compared to other existing well performing dynamic BN approaches in terms of recovering true relationships among genes. In addition, it provides a balance between searching for optimal networks and keeping biologically relevant regulatory interactions among variables.

ADEpedia-on-OHDSI: A next generation pharmacovigilance signal detection platform using the OHDSI common data model

Supplementing the Spontaneous Reporting System (SRS) with Electronic Health Record (EHR) data for adverse drug reaction detection could augment sample size, increase population heterogeneity and cross-validate results for pharmacovigilance research. The difference in the underlying data structures and terminologies between SRS and EHR data presents challenges when attempting to integrate the two into a single database. The Observational Health Data Sciences and Informatics (OHDSI) collaboration provides a Common Data Model (CDM) for organizing and standardizing EHR data to support large-scale observational studies. The objective of the study is to develop and evaluate an informatics platform known as ADEpedia-on-OHDSI, where spontaneous reporting data from FDA's Adverse Event Reporting System (FAERS) is converted into the OHDSI CDM format towards building a next generation pharmacovigilance signal detection platform. An extraction, transformation and loading (ETL) tool was designed, developed, and implemented to convert FAERS data into the OHDSI CDM format. A comprehensive evaluation, including overall ETL evaluation, mapping quality evaluation of drug names to RxNorm, and an evaluation of transformation and imputation quality, was then performed to assess the mapping accuracy and information loss using the FAERS data collected between 2012 and 2017. Previously published findings related to vascular safety profile of triptans were validated using ADEpedia-on-OHDSI in pharmacovigilance research. For the triptan-related vascular event detection, signals were detected by Reporting Odds Ratio (ROR) in high-level group terms (HLGT) level, high-level terms (HLT) level and preferred term (PT) level using the original FAERS data and CDM-based FAERS respectively. In addition, six standardized MedDRA queries (SMQs) related to vascular events were applied. A total of 4,619,362 adverse event cases were loaded into 8 tables in the OHDSI CDM. For drug name mapping, 93.9% records and 47.0% unique names were matched with RxNorm codes. Mapping accuracy of drug names was 96% based on a manual verification of randomly sampled 500 unique mappings. Information loss evaluation showed that more than 93% of the data is loaded into the OHDSI CDM for most fields, with the exception of drug route data (66%). The replication study detected 5, 18, 47 and 6, 18, 50 triptan-related vascular event signals in MedDRA HLGT level, HLT level, and PT level for the original FAERS data and CDM-based FAERS respectively. The signal detection scores of six standardized MedDRA queries (SMQs) of vascular events in the raw data study were found to be lower than those scores in the CDM study. The outcome of this work would facilitate seamless integration and combined analyses of both SRS and EHR data for pharmacovigilance in ADEpedia-on-OHDSI, our platform for next generation pharmacovigilance.

Matrix recovery with implicitly low-rank data

In this paper, we study the problem of matrix recovery, which aims to restore a target matrix of authentic samples from grossly corrupted observations. Most of the existing methods, such as the well-known Robust Principal Component Analysis (RPCA), assume that the target matrix we wish to recover is low-rank. However, the underlying data structure is often non-linear in practice, therefore the low-rankness assumption could be violated. To tackle this issue, we propose a novel method for matrix recovery in this paper, which could well handle the case where the target matrix is low-rank in an implicit feature space but high-rank or even full-rank in its original form. Namely, our method pursues the low-rank structure of the target matrix in an implicit feature space. By making use of the specifics of an accelerated proximal gradient based optimization algorithm, the proposed method could recover the target matrix with non-linear structures from its corrupted version. Comprehensive experiments on both synthetic and real datasets demonstrate the superiority of our method.

A Data Structure Perspective to the RDD-based Apriori Algorithm on Spark

During the recent years, a number of efficient and scalable frequent itemset mining algorithms for big data analytics have been proposed by many researchers. Initially, MapReduce-based frequent itemset mining algorithms on Hadoop cluster were proposed. Although, Hadoop has been developed as a cluster computing system for handling and processing big data, but the performance of Hadoop does not meet the expectation for the iterative algorithms of data mining, due to its high I/O, and writing and then reading intermediate results in the disk. Consequently, Spark has been developed as another cluster computing infrastructure which is much faster than Hadoop due to its in-memory computation. It is highly suitable for iterative algorithms and supports batch, interactive, iterative, and stream processing of data. Many frequent itemset mining algorithms have been re-designed on the Spark, and most of them are Apriori-based. All these Spark-based Apriori algorithms use Hash Tree as the underlying data structure. This paper investigates the efficiency of various data structures for the Spark-based Apriori. Although, the data structure perspective has been investigated previously, but for MapReduce-based Apriori, and it must be re-investigated in the distributed computing environment of Spark. The considered underlying data structures are Hash Tree, Trie, and Hash Table Trie. The experimental results on the benchmark datasets show that the performance of Spark-based Apriori with Trie and Hash Table Trie are almost similar but both perform many times better than Hash Tree in the distributed computing environment of Spark.