Relational Features Research Articles

Multimodal feature fusion based on object relation for video captioning

AbstractVideo captioning aims at automatically generating a natural language caption to describe the content of a video. However, most of the existing methods in the video captioning task ignore the relationship between objects in the video and the correlation between multimodal features, and they also ignore the effect of caption length on the task. This study proposes a novel video captioning framework (ORMF) based on the object relation graph and multimodal feature fusion. ORMF uses the similarity and Spatio‐temporal relationship of objects in video to construct object relation features graph and introduce graph convolution network (GCN) to encode the object relation. At the same time, ORMF also constructs a multimodal features fusion network to learn the relationship between different modal features. The multimodal feature fusion network is used to fuse the features of different modals. Furthermore, the proposed model calculates the length loss of the caption, making the caption get richer information. The experimental results on two public datasets (Microsoft video captioning corpus [MSVD] and Microsoft research‐video to text [MSR‐VTT]) demonstrate the effectiveness of our method.

Evaluation and prediction of the connect index between the Belt and Road countries and China based on the DANP method.

Interconnection is the priority direction of the Belt and Road initiative, which can provide substantial assistance to win-win cooperation. This study establishes a new indicator system from the five dimensions of policy, infrastructure, trade, finance, and people-to-people, evaluates the connect index of 63 Belt and Road countries from 2013 to 2020 based on the DEMATEL-ANP method which removes the potential subjective interference and interaction between indicators, and predicts the trend of the connect index by using the grey model. The findings indicate that the five dimensions of the Belt and Road connectivity have unevenly developed, among which the policy coordination has achieved the least. Singapore, Russia, and Malaysia have the highest connect index, and we can find that the 10 countries with the highest connect index are basically from East Asia & Pacific and Europe & Central Asia, which possess large economic and geographical differences. Moreover, there are 17 "omission areas" characterized by low national income, poor infrastructure, low population density, and small land areas along the Belt and Road. Finally, the Silk Road Economic Belt is facing structural imbalances in connectivity, and the relation features "proximity but not affinity" between China and its neighboring countries. These conclusions are friendly cautions and have constructive policy implications for the Belt and Road countries to achieve high-quality interconnection.

Exploring Privileged Features for Relation Extraction with Contrastive Student-Teacher Learning

Significant progress has been made by joint entity and relation extraction methods, which directly generate the relation triplets and mitigate the issue of overlapping relations. However, previous models generate the entity-relation triplets solely from input sentences. Such information is insufficient to support the modeling of interactive information between entities and relations. In this paper, we define the features that provide mutual supports for entity and relation detection but can only be accessed at training time as privileged features for relation extraction, and devise two teacher models to exploit privileged entity and relation features, respectively. Meanwhile, we propose a novel contrastive student-teacher learning framework for joint extraction of entities and relations (STER), where a student network is encouraged to amalgamate privileged knowledge from two expert teacher networks that additionally utilize the privileged features, based on contrastive learning. Experiment results on three benchmark datasets (i.e., ADE, SciERC and CoNLL04) demonstrate that STER has robust superiority over competitors and sets state-of-the-art. For reproducibility, we will release the data and source code once the paper is accepted.

Логико-лингвистическое моделирование угроз безопасности информации в информационных системах

Purpose: assessment of the possibility, definition of conditions and a brief description of the relational languages of logical-linguistic modeling for a formalized description and presentation of the processes of implementing information security threats in information systems. Method: application of the logical-linguistic modeling apparatus, which makes it possible to formally describe information security threats and a set of actions performed in the course of their implementation, taking into account the capabilities of relational description languages, such as Codd’s language, context-free plex-language, RX-code language, syntagmatic chains and semantic networks. Result: a brief description and comparative analysis of relational description languages and features that affect the possibility of their use for describing threats to information security and logical-linguistic modeling of their implementation processes are given. The expediency of such modeling is shown when creating promising expert systems designed for automated and automatic analysis of threats, when maintaining a data bank of threats based on the results of monitoring publications about them on the Internet. Examples of constructing formal logical-linguistic descriptions of well-known threats of computer attacks on information systems using RX-code languages and semantic networks are given, proposals are made for expanding the language of semantic networks to describe threats, taking into account new data on threats and methods for their implementation. It is noted that the proposed approach to modeling the processes of implementation of information security threats, as a rule, is applicable in the absence of the need to take into account the time factor when assessing the possibilities of their implementation.

Age-Based Automatic Voice Conversion Using Blood Relation for Voice Impaired

The present work presents a statistical method to translate human voices across age groups, based on commonalities in voices of blood relations. The age-translated voices have been naturalized extracting the blood relation features e.g., pitch, duration, energy, using Mel Frequency Cepstrum Coefficients (MFCC), for social compatibility of the voice-impaired. The system has been demonstrated using standard English and an Indian language. The voice samples for resynthesis were derived from 12 families, with member ages ranging from 8–80 years. The voice-age translation, performed using the Pitch synchronous overlap and add (PSOLA) approach, by modulation of extracted voice features, was validated by perception test. The translated and resynthesized voices were correlated using Linde, Buzo, Gray (LBG), and Kekre’s Fast Codebook generation (KFCG) algorithms. For translated voice targets, a strong (θ >~93% and θ >~96%) correlation was found with blood relatives, whereas, a weak (θ

Multiframe CenterNet Heatmap ROI Aggregation for Real-Time Video Object Detection

Though Two-stage video object detectors cannot perform detection in real time, the accuracy of them is normally higher than that of one-stage video object detectors. One essence is that two-stage video detectors can easily use feature information from adjacent frames to augment key frame features. How to extract and exploit temporal features in the video stream for one-stage detectors needs further exploration. CenterNet is an anchor-free one-stage object detector that regress bounding boxes from heatmap peaks. We propose to use detected peaks and regressed boxes which encompass peak points to determine the heatmap ROIs as the extracted object heatmap features. A new relation module is designed to evaluate the similarity of heatmap ROI features and output the relation features which can effectively augment the heatmap ROI features. In the video sequence the heatmap ROIs of multiple adjacent frames are aggregated to a heatmap ROI of the key frame. Compared to CenterNet and other CenterNet-based video object detectors, our method achieves improved online real-time performance on ImageNet VID dataset with 78.8% mAP at 36 FPS.

An Efficient Scheme for Data Pattern Matching in IoT Networks

The Internet has become an unavoidable trend of all things due to the rapid growth of networking technology, smart home technology encompasses a variety of sectors, including intelligent transportation, allowing users to communicate with anybody or any device at any time and from anywhere. However, most things are different now. Background: Structured data is a form of separated storage that slows down the rate at which everything is connected. Data pattern matching is commonly used in data connectivity and can help with the issues mentioned above. Aim: The present pattern matching system is ineffective due to the heterogeneity and rapid expansion of large IoT data. The method requires a lot of manual work and has a poor match with real-world applications. In the modern IoT context, solving the challenge of automatic pattern matching is complex. Methodology: A three-layer mapping matching is proposed for heterogeneous data from the IoT, and a hierarchical pattern matching technique. The feature classification matching, relational feature clustering matching, and mixed element matching are all examples of feature classification matching. Through layer-by-layer matching, the algorithm gradually narrows the matching space, improving matching quality, reducing the number of matching between components and the degree of manual participation, and producing a better automatic mode matching. Results: The algorithm's efficiency and performance are tested using a large number of data samples, and the results show that the technique is practical and effective. Conclusion: the proposed algorithm utilizes the instance information of the data pattern. It deploys three-layer mapping matching approach and mixed element matching and realizes the automatic pattern matching of heterogeneous data which reduces the matching space between elements in complex patterns. It improves the efficiency and accuracy of automatic matching.

A Subset Scaling Recursive Feature Collection Based DDoS Detection Using Behavioural Based Ideal Neural Network For Security In A Cloud Environment

The Cloud Computing model improves cloud resources and reduces cloud user latency. The Cloud Computing model expands services such as network equipment, computer capabilities and storage devices. Cloud services are distributed naturally so that millions of users can share. Because of this, the cloud environment has many security tasks. Distributed Denial of Service (DDoS) Attacks and Techniques for Detecting and Preventing analysis in a cloud computing environment. Previous analysis has some drawbacks in DDoS attack detection, including security issues, Low Accuracy, and data loss. Identifying a DDoS attack is very difficult because it is a computational problem that needs to be addressed. To overcome the issues, this work proposed the method, Subset Scaling Recursive Factor Feature selection (S2RF2S), used to detect DDoS attacks based on Lattice Structural access rate using Soft-Max Behavioral Based Ideal Neural Network (SxB2IN2) used to detect DDoS attack detection. Initially, using the collection of the dataset for analysis in preprocessing step and reducing the imbalanced or irrelevant data from the dataset. Then, Subset Scaling Recursive Factor Feature selection (S2RF2S) for filtering the relational features based on the Lattice structural access rate. The lack of traffic bandwidth aspect balances; Social Spider Optimization analyzes these mutual balances to select Attack Features (S2OSAF) using features based on each feature's weights. Soft-Max activation for creating neurons to evaluate the features into subgroup feature selection and training with Behavioral Based Ideal Neural Network (SxB2IN2). This proposed system performs better for data loss and detecting DDoS attacks. The simulation results show the performance of the proposed method to avoid security issues in Cloud Computing.

Counterfactual inference to predict causal knowledge graph for relational transfer learning by assimilating expert knowledge --Relational feature transfer learning algorithm

Transfer learning (TL) is a machine learning (ML) method in which knowledge is transferred from the existing models of related problems to the model for solving the problem at hand. Relational TL enables the ML models to transfer the relationship networks from one domain to another. However, it has two critical issues. One is determining the proper way of extracting and expressing relationships among data features in the source domain such that the relationships can be transferred to the target domain. The other is how to do the transfer procedure. Knowledge graphs (KGs) are knowledge bases that use data and logic to graph-structured information; they are helpful tools for dealing with the first issue. The proposed relational feature transfer learning algorithm (RF-TL) embodies an extended structural equation modelling (SEM) as a method for constructing KGs. Additionally, in fields such as medicine, economics, and law related to people’s lives and property safety and security, the knowledge of domain experts is a gold standard. This paper introduces the causal analysis and counterfactual inference in the TL domain that directs the transfer procedure. Different from traditional feature-based TL algorithms like transfer component analysis (TCA) and CORelation Alignment (CORAL), RF-TL not only considers relations between feature items but also utilizes causality knowledge, enabling it to perform well in practical cases. The algorithm was tested on two different healthcare-related datasets — sleep apnea questionnaire study data and COVID-19 case data on ICU admission — and compared its performance with TCA and CORAL. The experimental results show that RF-TL can generate better transferred models that give more accurate predictions with fewer input features.

Super-resolution semantic segmentation with relation calibrating network

To achieve high-resolution segmentation results, typical semantic segmentation models often require high-resolution inputs. However, high-resolution inputs inevitably bring high cost on computation, which limits its application seriously in realistic scenarios. To address the problem, we propose to predict a high-resolution semantic segmentation result with a degraded low-resolution image as input, which is called super-resolution semantic segmentation in this paper. We further propose a Relation Calibrating Network (RCNet) for this task. Specifically, we propose two modules, namely Relation Upsampling Module (RUM) and Feature Calibrating Module (FCM). In RUM, the input feature map generates the relation map of pixels in low-resolution, which is then gradually upsampled to high-resolution. Meanwhile, FCM takes the input feature map and the relation map from RUM as inputs, gradually calibrating the feature. Finally, the last FCM outputs the high-resolution segmentation results. We conduct extensive experiments to verify the effectiveness of our method. Specially, we achieve a comparable segmentation result (from 70.01% to 70.90%) with only 1/4 of the computational cost (from 1107.57 to 255.72 GFLOPs) based on FCN on Cityscapes dataset.

A Scoring Procedure for Malignant Narcissism Based on Personality Inventory for DSM-5 Facets

The current study focuses on the development and validation of a scoring procedure for malignant narcissism using the Personality Inventory for DSM-5, a self-report measure of Criterion B from the Alternative Model for Personality Disorders. In Study 1, a prototype matching approach was used to aggregate ratings from 15 clinicians specializing in personality disorder treatment and/or assessment. Indices of inter-rater agreement and inter-rater reliability revealed high convergence as to the most important maladaptive facets for malignant narcissism. The scoring procedure, based on additive counts for score computation, included eleven Criterion B facets covering core features of malignant narcissism. Study 2 evaluated the criterion and incremental validity of the scoring procedure in a sample of 288 patients from a personality disorder treatment clinic, as well as in a sample of 1103 participants from the community. In both samples, results from nonparametric mean comparisons, receiver operating characteristic curves, bivariate Pearson correlations, and hierarchical multiple linear regressions showed significant associations between malignant narcissism and broader components of personality functioning, as well as with relevant emotional, relational, and/or behavioral features. This new scoring procedure is a simple and valid method for measuring malignant narcissism, and is suitable for clinical and research settings.

One-Shot Relation Learning for Knowledge Graphs via Neighborhood Aggregation and Paths Encoding

The relation learning between two entities is an essential task in knowledge graph (KG) completion that has received much attention recently. Previous work almost exclusively focused on relations widely seen in the original KGs, which means that enough training data are available for modeling. However, long-tail relations that only show in a few triples are actually much more common in practical KGs. Without sufficiently large training data, the performance of existing models on predicting long-tail relations drops impressively. This work aims to predict the relation under a challenging setting where only one instance is available for training. We propose a path-based one-shot relation prediction framework, which can extract neighborhood information of an entity based on the relation query attention mechanism to learn transferable knowledge among the same relation. Simultaneously, to reduce the impact of long-tail entities on relation prediction, we selectively fuse path information between entity pairs as auxiliary information of relation features. Experiments in three one-shot relation learning datasets show that our proposed framework substantially outperforms existing models on one-shot link prediction and relation prediction.

Application of Tamm Plasmon Polaritons and Cavity Modes for Biosensing in the Combined Spectroscopic Ellipsometry and Quartz Crystal Microbalance Method.

Low-cost 1D plasmonic photonic structures supporting Tamm plasmon polaritons and cavity modes were employed for optical signal enhancement, modifying the commercially available quartz crystal microbalance with dissipation (QCM-D) sensor chip in a combinatorial spectroscopic ellipsometry and quartz microbalance method. The Tamm plasmon optical state and cavity mode (CM) for the modified mQCM-D sample obtained sensitivity of ellipsometric parameters to RIU of ΨTPP = 126.78 RIU−1 and ΔTPP = 325 RIU−1, and ΨCM = 264 RIU−1 and ΔCM = 645 RIU−1, respectively. This study shows that Tamm plasmon and cavity modes exhibit about 23 and 49 times better performance of ellipsometric parameters, respectively, for refractive index sensing than standard spectroscopic ellipsometry on a QCM-D sensor chip. It should be noted that for the optical biosensing signal readout, the sensitivity of Tamm plasmon polaritons and cavity modes are comparable with and higher than the standard QCM-D sensor chip. The different origin of Tamm plasmon polaritons (TPP) and cavity mode (CM) provides further advances and can determine whether the surface (TPP) or bulk process (CM) is dominating. The dispersion relation feature of TPP, namely the direct excitation without an additional coupler, allows the possibility to enhance the optical signal on the sensing surface. To the best of our knowledge, this is the first study and application of the TPP and CM in the combinatorial SE-QCM-D method for the enhanced readout of ellipsometric parameters.

SRAI-LSTM: A Social Relation Attention-based Interaction-aware LSTM for human trajectory prediction

Pedestrian trajectory prediction is one of the important research topics in the field of computer vision and a key technology of autonomous driving system. Walking in groups is a common social behavior in which pedestrians pay more attention to the movements of their companions while walking. Motivated by this idea, we propose a Social Relation Attention-based Interaction-aware LSTM (SRAI-LSTM) to model this social behavior for trajectory prediction. We design a social relation encoder module to capture social relation feature between pedestrians through their relative positions. Afterwards, the social relation features are adopted to acquire social relation attentions among pedestrians. Social interaction modeling is achieved by utilizing social relation attentions to aggregate motion features from neighbor pedestrians. Experimental results on two public pedestrian trajectory datasets (ETH and UCY) demonstrate that our proposed model achieves superior performances compared with state-of-the-art methods on ADE and FDE metrics.

Perceptual Organisation Affects Perception and Appreciation of Abstract Art: A Case Study with Black Square and Red Square by Kazimir Malevich

Abstract Perceptual organisation is hypothesised as a key in the perception and appreciation of abstract art. Here, we investigated how relational and compositional features affected the perception and aesthetic appreciation of Black Square and Red Square by Kazimir Malevich (1915). We studied how (i) the presence and obliquity of the red square and (ii) the relative configuration of the black and red square affected the detectability of the obliquity of the black square in this artwork. Results showed that the simultaneous presence and obliquity of the red square masked the obliquity of the original black square. The likelihood of the black square being incorrectly perceived as an exact square was always maximum in the original configuration and even slight alterations in the original configuration of the work resulted in the obliquity of the black square to be noticed. The original artwork was more aesthetically preferred compared to its alternatives. We argue that the artist may have intentionally set the configuration to mask the obliquity of the black square and maximise the aesthetic preference.

Exploring the role of faculty and staff mentors in fostering ethical leadership among undergraduate students: “We have to narrow the distance”

PurposeThe purpose of this interdisciplinary study is to identify the aspects that faculty, student affairs educators and students indicate as salient for effective mentoring relationships that enhance ethical leadership development.Design/methodology/approachThis exploratory qualitative inquiry used the Relational-Ethical-Affective-Dialogic (READ) mentoring model as a framework to examine the experiences of 13 undergraduate mentees and faculty/staff mentors in a formal mentoring program. Each study participant engaged in one semi-structured interview. Researchers coded and analyzed data using the sort and sift, think and shift process identifying power quotes to guide the thematic analysis.FindingsThe data collected in this study revealed insights into the aspects of mentor relationships that both undergraduate mentees and their mentors perceived as contributing to students' ethical leadership development. Salient elements included: (1) relational features of the mentee-mentor dynamic including trust and reciprocity; (2) structural features of the mentoring program including its focus on ethics; and (3) mentoring approaches that were attentive to power and positionality within the mentoring relationship and involved professional judgment about self-disclosure.Originality/valueThis study adds to the literature by exploring effective mentoring for ethical leadership development across disciplines. With colleges and universities serving a vital role in preparing the next generation of leaders for ethical engagement in their democratic and professional roles after graduation, it is imperative to broaden our understanding of how faculty and staff can support students' ethical leadership development.

Maternal-infant bonding for the opioid exposed dyad: A rodgers' evolutionary concept analysis.

This concept analysis aimed to offer insight into how opioid exposure affects the quality and construction of the maternal-infant bond. Maternal-infant bonding can be influenced by maternal opioid use disorder and the neonatal intensive care unit environment. Many interventions for mothers with opioid use disorder focus on enhancing parental skills, but often the emotional and relational features of the maternal-infant bond have been overlooked. Literature from Cumulative Index to Nursing and Allied Health Literature (CINAHL), Embase, PubMed, and PsycINFO published from January 2011 to June 2021 using "attachment theory," "mother-infant attachment," "maternal-infant bonding," "neonatal opioid withdrawal syndrome," and "neonatal abstinence syndrome" as key terms. Rodgers' method of concept analysis was used to determine the antecedents, attributes, and consequences of maternal-infant bonding. The key attributes of bonding for the opioid-exposed mother-infant dyad are engaged mothering, recognition of risk, affection, and respectful maternity care. The antecedents are closeness, selflessness, and purposeful touch. The consequences are sobriety, custody, love, and security.

Generalized attention-weighted reinforcement learning

In neuroscience, attention has been shown to bidirectionally interact with reinforcement learning (RL) to reduce the dimensionality of task representations, restricting computations to relevant features. In machine learning, despite their popularity, attention mechanisms have seldom been administered to decision-making problems. Here, we leverage a theoretical model from computational neuroscience – the attention-weighted RL (AWRL), defining how humans identify task-relevant features (i.e., that allow value predictions) – to design an applied deep RL paradigm. We formally demonstrate that the conjunction of the self-attention mechanism, widely employed in machine learning, with value function approximation is a general formulation of the AWRL model. To evaluate our agent, we train it on three Atari tasks at different complexity levels, incorporating both task-relevant and irrelevant features. Because the model uses semantic observations, we can uncover not only which features the agent elects to base decisions on, but also how it chooses to compile more complex, relational features from simpler ones. We first show that performance depends in large part on the ability to compile new compound features, rather than mere focus on individual features. In line with neuroscience predictions, self-attention leads to high resiliency to noise (irrelevant features) compared to other benchmark models. Finally, we highlight the importance and separate contributions of both bottom-up and top-down attention in the learning process. Together, these results demonstrate the broader validity of the AWRL framework in complex task scenarios, and illustrate the benefits of a deeper integration between neuroscience-derived models and RL for decision making in machine learning.

Considering equity in global health collaborations: A qualitative study on experiences of equity.

International collaborations have become the standard model for global health research and often include researchers and institutions from high income countries (HICs) and low- and middle-income countries (LMICs). While such collaborations are important for generating new knowledge that will help address global health inequities, there is evidence to suggest that current forms of collaboration may reproduce unequal power relations. Therefore, we conducted a qualitative study with scientists, researchers and those involved in research management, working in international health collaborations. Interviews were conducted between October 2019 and March 2020. We conducted 13 interviews with 15 participants. From our findings, we derive three major themes. First, our results reflect characteristics of equitable, collaborative research relationships. Here we find both relational features, specifically trust and belonging, and structural features, including clear contractual agreements, capacity building, inclusive divisions of labour, and the involvement of local communities. Second, we discuss obstacles to develop equitable collaborations. These include exclusionary labour practices, donor-driven research agendas, overall research culture, lack of accountability and finally, the inadequate financing of indirect costs for LMIC institutions. Third, we discuss the responsibilities for promoting science equity of funders, LMIC researchers, LMIC institutions, and LMIC governments. While other empirical studies have suggested similar features of equity, our findings extend these features to include local communities as collaborators in research projects and not only as beneficiaries. We also suggest the importance of funders paying for indirect costs, without which the capacity of LMIC institutions will continually erode. And finally, our study shows the responsibilities of LMIC actors in developing equitable collaborations, which have largely been absent from the literature.

Multi-Graph Cooperative Learning Towards Distant Supervised Relation Extraction

The Graph Convolutional Network (GCN) is a universal relation extraction method that can predict relations of entity pairs by capturing sentences’ syntactic features. However, existing GCN methods often use dependency parsing to generate graph matrices and learn syntactic features. The quality of the dependency parsing will directly affect the accuracy of the graph matrix and change the whole GCN’s performance. Because of the influence of noisy words and sentence length in the distant supervised dataset, using dependency parsing on sentences causes errors and leads to unreliable information. Therefore, it is difficult to obtain credible graph matrices and relational features for some special sentences. In this article, we present a Multi-Graph Cooperative Learning model (MGCL), which focuses on extracting the reliable syntactic features of relations by different graphs and harnessing them to improve the representations of sentences. We conduct experiments on a widely used real-world dataset, and the experimental results show that our model achieves the state-of-the-art performance of relation extraction.