17837 research outputs found
Sort by
Enhancing Quality-Diversity Optimization Through Domain-Specific Dissimilarity as Crowding Distance
Get PDFQuality-diversity algorithms aim to simultaneously optimize solution performance and maintain diversity within a population. In this paper, we explore the use of NSGA-II as a quality-diversity algorithm for the evolutionary design of 3D structures, modifying its crowding distance calculation to utilize dissimilarity measures. While NSGA-II is widely employed for multi-objective optimization, its use of fitness for calculating crowding distance may not be the most effective for tasks requiring solution diversity. We propose leveraging both genetic and phenotypic dissimilarity metrics to improve diversity management. To evaluate this approach, we compare the standard NSGA-II using fitness-based crowding distance and Diversity-Enhancing NSGA-II (DE-NSGA-II) using various combinations of dissimilarity-based metrics for crowding distance and diversity scores. Experiments are conducted using two distinct genetic representations on two optimization tasks: height of the center of gravity of passive structures and velocity of active structures. Results demonstrate the potential of dissimilarity-based crowding distance to enhance the diversity and overall quality of solutions in complex evolutionary design tasks
A Novel Lift Adjustment Methodology for Improving Association Rule Interpretation
Get PDFAssociation rules can offer a human-interpretable insight extracted from data. The lift measures used for evaluating association rules in classical Association Rule Mining (ARM) contexts are mainly based on traditional and well-known ones but suffer from interpretation inadequacy when dealing with skewed distributions or low support. This study introduces a new lift adjustment approach with four methods to overcome traditional lift measures and identify the best rules in association rule mining. More concretely, our main objective is to improve the interpretability of association rules to make them more practically relevant for decision-making. We propose an approach incorporating four novel lift adjustment methods (smoothed, weighted, log, and threshold-adjusted lift) to achieve this. We introduce a flexible, dynamic approach combined with four new lift adjustment methods: smoothed, weighted, logarithm, and threshold-adjusted lift. Each technique addresses specific limitations of the traditional lift measure and better captures the reliable representation of item associations by exaggerating stronger relationships or smoothing weaker ones. The proposed methods applied context-aware rule evaluation and adjustment based on measures of relative significance (e.g., Jaccard similarity). The experimental results involving real-world data and synthetic datasets reveal new methods’ effectiveness and robustness in understanding the strengths of association rules and provide a comprehensive view that considers item importance. We evaluate the performance stability of our proposed methods using statistical analysis, including ANOVA, chi-squared, t-tests, and effect size metrics
VIA-SLAM:An Underwater Visual–Inertial–Acoustic SLAM With Integrated DVL
Get PDFArticle proposes a visual-inertial–acoustic SLAM (VIA-SLAM) for the localization of unmanned underwater vehicles (UUVs). The initialization of SLAM is a difficult task because of the slow-motion characteristics of UUVs and the challenging underwater environment with weak textures and low light conditions. The SLAM system integrates a Doppler velocity log (DVL), and we propose a DVL-assisted initialization method to enhance the accuracy of system initialization. The velocity measurements from the DVL are used to construct the DVL residuals, which is jointly optimized with the reprojection residuals and the inertial residuals to enhance the localization accuracy of the system. In addition, the reliability of visual information is affected by the complex underwater environment. To address the problem of decreased localization accuracy due to poor image quality, we propose an adaptive dynamic weight factor to assess the impact of image quality on localization performance and accordingly adjust the weight of visual information. Finally, we evaluate the proposed VIA-SLAM using the open underwater datasets and the data collected from actual lake trials. The experimental results show that our SLAM system significantly improves localization accuracy compared with the state-of-the-art methods
Hyperbolic Adversarial Learning for Personalized Item Recommendation
Get PDFPersonalized recommendation systems are indispensable intelligent components for social media and e-commerce. Traditional personalized item recommendation models are vulnerable to adversarial perturbations, resulting in poor robustness. Although adversarial learning-based recommendation models are able to improve the robustness, they inherently model the interaction relationships between users and items in Euclidean space, where it is difficult for them to capture the hierarchical relationships among entities. To address the above issues, we propose a hyperbolic adversarial learning based personalized item recommendation model, called HALRec. Specifically, HALRec models the interactions in hyperbolic space and utilizes hyperbolic distances to measure the similarities among entities. Moreover, instead of in Euclidean space, HALRec exploits the adversarial learning technique in hyperbolic space, i.e., HAL-Rec maximizes the hyperbolic adversarial perturbations loss while minimizing the hyperbolic based Bayesian personalized ranking loss. Hence, HALRec inherits the advantages of hyperbolic representation learning in capturing hierarchical relationships and adversarial learning in enhancing the robustness of the recommendation model. In addition, we utilize tangent space optimization to simplify the learning of model parameters. Experimental results on real-world datasets show that our proposed hyperbolic adversarial learning-based personalized item recommendation method outperforms the state-of-the-art personalized recommendation algorithms
Similarities and Differences between Terrorists and their Supporters:Findings from Northern Ireland
Get PDFWhy do only a few people radicalise when many appear to have been exposed to potential causes of radicalisation? This issue is referred to as the specificity problem and is widely recognised as one of the fundamental questions facing terrorism research. Also recognised is that meaningful answers will only be obtained through comparative research using control groups who share many of the same traits, characteristics and contexts of the terrorists, but who did not progress to involvement in terrorism. The current study was conducted in Northern Ireland and compares 17 former paramilitary members with a control group of 12 paramilitary sympathizers using a structured survey. 22 variables connected to radicalisation were examined. Significant differences were found between the two groups on four of these variables. The findings are discussed in relation to the specificity problem and the wider literature on radicalisation
Digital CBT for OCD: Perspectives on the therapeutic alliance from adolescents and therapists
Get PDFAdolescents with obsessive-compulsive disorder (OCD) face unique challenges in engaging with cognitive-behavioural therapy (CBT), particularly the distress associated with exposure and response prevention (ERP) tasks. As digital delivery of CBT becomes increasingly widespread, questions remain about how the therapeutic alliance - a key predictor of treatment outcomes - is developed and sustained in this format. This study explored how adolescents and therapists experience the process of building and maintaining the therapeutic alliance during digital CBT for OCD, with particular attention to barriers, facilitators, and adaptations. Semi-structured interviews were conducted with six therapists and four adolescents who had participated in digital CBT for OCD. Although recruitment targeted ‘adolescents’, the client participants predominantly represented the upper end of adolescence, with all aged 18 or over (i.e. aligning with many definitions of emerging adulthood). For continuity with the study framing and service context, the term ‘adolescent’ is retained to denote older adolescents/emerging adults and interpret the findings with this developmental positioning in mind. A reflexive thematic analysis was used to identify patterns in participants’ experiences, highlighting both shared and divergent perspectives across the two groups. The analysis identified seven themes, illustrating both the opportunities and obstacles when fostering the alliance in digital settings. Therapists described making deliberate adaptations to humanise the digital space and maintain emotional presence, while adolescents emphasised the importance of privacy and developmentally-attuned approaches to support engagement. Technical and home-based disruptions, and the emotional distance of digital therapy were reported as barriers to openness - particularly during ERP tasks that rely on trust and in-the-moment support. Collaboration and flexibility emerged as essential strategies for overcoming these challenges, with both groups emphasising the importance of tailoring therapy to adolescents’ individual needs and daily environments. The findings underscore the importance of therapist adaptability, collaborative planning, and intentional rapport-building in digital CBT for OCD. Collectively, the themes point to concrete strategies - such as normalising the digital format, co-designing ERP tasks, and protecting privacy - that therapists can integrate into sessions to foster adolescent engagement and relational depth. Embedding these competencies into therapist training and service protocols may enhance digital CBT delivery and thus warrants systematic evaluation in future research
Data sources:
Access Repository Dashboard
Do you manage Royal Holloway - Pure? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!