116320 research outputs found
Sort by
Workplace learning and Complexity Theory: the telos of small groups
No full textWorkplace learning has outgrown its theoretical background in Human Capital Theory. After showing how and why this welcome change has occurred, this chapter draws upon Complexity Theory to demonstrate the distinctive, but powerful, learning that small groups (2–12 participants) can generate in workplaces. Teams are prominent and common examples of this. Complexity Theory puts relationality (not things or individuals) first. It supports learning from holistic human experience (not merely the cognitive or social), and emphasises the 'telos' or purposes of small-group work (such as completing a project, or the wellness of the patient). Two real-life case studies, involving shipping navigation and 'Titan', a Japanese manufacturing company, illustrate how Complexity Theory sheds new light on how learning occurs within workplace practices. We claim that small groups should be the heart of understanding workplace learning because new knowledge emerges through participants' shared activities, for the benefit of all involved. This is distinctively different from any knowledge individuals can accrue by themselves
Causal cascading convolution networks for multi-behavior sequential recommendation
No full textExploring multi-behavioral sequence recommendation has emerged as a crucial topic in recent years. It is well-known that user interactions on online platforms, such as social media websites and news aggregation applications, revolve around singular actions like reading or clicking. These interactions may also involve diverse behaviors, such as commenting, sharing, and bookmarking. Each of these varied behaviors reflects different facets of user preferences in their interaction sequences with items. Consequently, understanding and combining these diverse behaviors to effectively represent user preferences becomes vital. Most existing methods construct user preferences and interests based on correlations among users, items, and behaviors. However, in real-world scenarios, causality often drives users to make their next decision rather than merely relying on correlation. Unfortunately, this causal relationship is frequently overlooked by most multi-behavior models. To address this gap, we propose a Causality-based Multi-behavior Sequential Recommendation (CMSR) framework to capture the underlying causal relationships among user behaviors and predict future actions. Specifically, CMSR first independently encodes each behavioral sequence to capture user preferences across different behaviors. It then aggregates inter-item behavioral relationships through hypergraph convolution. We also employ cascade networks to capture directional dependencies in multi-behavior sequences within the behavior chain. Finally, CMSR transfers the influences of causal relationships among behaviors by utilizing a causal graph construction approach. To assess the efficacy of the proposed CMSR model, a series of comprehensive experiments were carried out utilizing a pair of datasets derived from actual operational environments. The outcomes of these experiments illustrate the CMSR's effectiveness and its superior performance relative to the established baseline techniques
High-precision inversion of vegetation parameters in the AI era: Integrating hyperspectral remote sensing and deep learning.
Full text linkSelf-supervised 3D Reconstruction of Tibia and Fibula from Biplanar X-rays
No full textWith the growing number of patients experiencing knee-related conditions, total knee arthroplasty (TKA) has become a common procedure, where a 3D visualisation of the patient’s tibia and fibula is essential for preoperative planning. Traditional imaging techniques, such as computed tomography (CT), often expose patients to high levels of radiation or impose significant financial costs. As an alternative, this paper proposes a novel approach that reconstructs a 3D model of the tibia and fibula using only two X-ray images (taken from the coronal and sagittal planes) and a general template, significantly reducing radiation exposure and financial burden. Our algorithm of 3D reconstruction for patient-specific anatomies combines point-based deformation with deep learning techniques. Initially, the general model undergoes a preliminary deformation to match the patient tibia and fibula dimensions. This pre-deformed model then serves as a template, followed by a fine deformation process via a self-supervised graph convolutional network (GCN), whose parameters are trained iteratively by comparing the template projection and the X-ray measurements. Following tests in simulations, cadaver experiments, and in-vivo experiments, our proposed algorithm demonstrates state-of-the-art accuracy and exceptional robustness across different evaluation metrics. Our code is available at https://github.com/DrKaiPan/tfDeform_GCN.gi
Green spaces
No full textThis is a definition of green spaces in the Thematic Encyclopedia of Regional Science. This thematic encyclopedia explores the multifaceted world of regional science, presenting a systematic and coherent overview of its central topics. It highlights the interdisciplinary nature of the field, examining the wide range of concepts, theories, methods, and models that shape spatial-oriented approaches to the social sciences. Contributions from expert scholars delve into key aspects of regional science, from urban poverty and natural resource management to smart cities and AI. Highly accessible entries cover the definition, history, theoretical background, and applications of each topic, as well as avenues for future research
Next generation ESKAPE-E superbugs: identifying transmissible locus of stress tolerance and antibiotic resistance in pandemic bacterial lineages.
Full text linkThe transmissible locus of stress tolerance (tLST) confers enhanced resilience to stressors, including chlorine-based disinfectants and heat. A systematic examination of 48,183 complete bacterial genomes representing 7,190 unique species identified tLST, including six novel variants, in 2,128 genomes spanning 46 bacterial species across 10 families. Among the analysed sequences, tLST was most common in ESKAPE-E bacteria, including highly drug-resistant pandemic lineages of Pseudomonas aeruginosa ST111 and Klebsiella pneumoniae ST20. The tLST1_AW1.7 variant was predominantly associated with F plasmids and was the most common variant in Enterobacter and Klebsiella, whereas tLST1 was near exclusive to E. coli, and tLST3a was dominant in Pseudomonas. This comprehensive mapping of tLST distribution highlights its potential significance in bacterial adaptation and persistence across diverse ecological and anthropogenically impacted niches, and has implications for infection control and eradication, waste and drinking water management, food animal and fresh produce production, and food safety
Benchmark-neutral pricing
Full text linkThe paper proposes benchmark-neutral pricing and hedging for long-term contingent claims. It employs the growth optimal portfolio of the stocks as numéraire and the new benchmark-neutral pricing measure for pricing. For the assumed ‘natural’ dynamics of a well-diversified stock portfolio, which are those of the continuous limit of a branching process of diversified wealth in some activity time, this pricing measure turns out to be an equivalent probability measure. This is not the case for the putative risk-neutral pricing measure. Benchmark-neutral pricing identifies the minimal possible prices of contingent claims. Risk-neutral prices of long-term contracts can be significantly more expensive than necessary. The extremely accurate hedge of a long-term zero-coupon bond illustrates the proposed pricing and hedging method
Revealing vulnerable regions through diverse adversarial examples
Full text linkCurrent explainable AI approaches to Deep Neural Networks (DNNs) primarily aim to understand network behavior by identifying key input features that influence predictions. However, these methods often fail to identify vulnerable regions in the input that are sensitive to minor perturbations and pose significant security risks. The vulnerability of DNNs is typically studied through adversarial examples, but traditional norm-based algorithms, lacking spatial constraints, distribute perturbations across the entire image, obscuring these critical areas. To overcome this limitation, we propose the Vulnerable Region Discovery Attack (VrdAttack), an efficient method that leverages Differential Evolution to generate diverse one-pixel perturbations, enabling the discovery of vulnerable regions and uncovering pixel-level vulnerabilities in Deep Neural Networks (DNNs). Our extensive experiments on CIFAR-10 and ImageNet demonstrate that our proposed VrdAttack outperforms existing methods in identifying diverse critical weak points in an input, highlighting model-specific vulnerabilities, and revealing the impact of adversarial training on these vulnerable regions
Potential confounding mutations in Keio knockout strains: implications for research accuracy.
Full text linkThe Keio library of single-gene knock-out mutants of Escherichia coli is useful for the research community. It has been used to analyze the role of various E. coli genes in alcohol tolerance, multi-drug resistance, and biofilm formation. The current study provides a general overview of potential single nucleotide polymorphisms (SNPs), insertion-deletion of bases (≤50 nucleotides, INDELs) in the genome of a set of 21 knock-out mutants of the Keio collection in comparison to the parent strain. A small number of SNPs and INDELs were predicted in the coding and intergenic regions of the knock-out mutants in comparison to the parental strain through sequencing and bioinformatic analysis. Mutations in the coding regions of genes (different from the actual gene knocked out in the mutants) led to different types of mutations in the affected genes, ranging from nonsense mutations to frameshift mutations, which could affect the functionality of the resulting gene products. These mutations in the intergenic and coding regions could lead to phenotypic differences in the single-gene knock-out mutant strains in comparison to the parent strain, independent of the desired gene deletion. This, in turn, could be misinterpreted by researchers using these strains as differences caused by the missing gene. While this is a preliminary study based on only 21 strains of the Keio collection, the deleted genes in the mutants used in this study were approximately evenly distributed across the entire genome. This study likely indicates the possibility of such mutations in other Keio strains, although a larger sample size of knock-out mutants would be required to understand the likelihood of such mutations across the library.IMPORTANCEThe Keio library of single-gene knock-out mutants of Escherichia coli has been widely used for a variety of studies. However, mutations might appear in the genome of these strains over time, leading to differences in the characteristics of the mutant and parent strains that are independent of the gene deletions of interest. This study predicts the presence of a few SNPs and INDELs in some of the knock-out mutants from the Keio collection, which could potentially alter the phenotypic attributes of the knock-out mutants with no role of the deleted gene towards this change. Therefore, this study highlights the possibility of the presence of such mutations in other strains of the library and the importance of conducting additional steps, such as complementation assays, to confirm the outcomes of studies comparing specific attributes of the knock-out mutants with the parental strain
Multi-class fruit ripeness detection using YOLO and SSD object detection models
Full text linkAccurate fruit ripeness detection is critical to reducing post-harvest losses and improving quality control in agricultural systems. This study benchmarks four object detection models—YOLOv5, YOLOv6, YOLOv7, and SSD-MobileNetv1—for multi-class ripeness classification of strawberries and avocados across four stages: unripe, partially ripe, ripe, and rotten. The dataset, captured under natural conditions, has been manually annotated and published for public access. YOLOv6 achieved the highest mean Average Precision (99.5%) and demonstrated a strong balance between accuracy and real-time inference speed (85.2 FPS). All models were evaluated using standard classification metrics and cross-validated through a 5-fold approach to ensure robustness. The results indicate YOLOv6 as the most reliable model for smart fruit sorting and quality monitoring applications. This study offers a reproducible benchmarking pipeline and contributes toward the development of deployable deep learning solutions in precision agriculture
Data sources:
Access Repository Dashboard
Do you manage OPUS? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!