28579 research outputs found
Sort by
Taking Variance Seriously: Visualizing the Statistical and Substantive Significance of ARCH-GARCH Models
Full text linkAutoregressive Conditional Heteroskedasticity (ARCH) and Generalized ARCH (GARCH) models allow users to estimate the conditional mean and conditional error variance of a time series process. While simulation methods exist to disaggregate the short- and long-run effects of covariate shocks to the conditional mean, scholars' inferences about the conditional error variance are currently limited to tabular interpretation. We propose a novel method of interpretation that moves beyond these tabular inferences. First, we show how changes in ARCH-GARCH processes are conditional on starting values, other covariates, and dynamics, which has led to incomplete or even incorrect inferences. We then develop three bootstrapping techniques to simulate conditional error variance model results and showcase the usefulness of each through replication of prominent studies. Our techniques demonstrate the crucial role of simulation and prediction for drawing statistical and substantive inferences about the volatility of dynamic time series processes
Landscape of nucleosome repositioning in glioblastoma multiforme
No full textGlioblastoma (GBM) is classified as a grade IV astrocytoma, characterised by its aggressive nature and poor prognosis. It represents a major problem in oncology given its abysmal survival rate and lack of effective treatment options to cure the malignancy and prevent relapses. The current methods for diagnostics generally include magnetic resonance imaging (MRI) followed by histopathological confirmation via analysis of tissue biopsy, which is an invasive method. Furthermore, such diagnostics methods often result in brain cancer detection at late stages. One alternative method is liquid biopsy, which uses bodily fluids, e.g. blood plasma or cerebral spinal fluid (CSF), to extract and analyse cell-free DNA (cfDNA) to detect genomic or epigenomic alterations characteristic of the disease. The aim of this project was to analyse a specific type of change in cfDNA reflecting nucleosome positioning in the cells of origin, which may be used for GBM diagnostics. Using data from a cohort of 10 GBM patients, I have analysed nucleosome positioning in paired normal/tumour brain tissues and the corresponding cfDNA. I compared nucleosome profiles from these three sources around binding sites of transcription factors (TFs) and developed new ways of data normalisation and visualisation with heatmaps and clustering. This work provided a shortlist of TFs that can be used as biomarkers and highlighted the challenges encountered in liquid biopsies of GBM and suggested new approaches to overcome these
Lightweight Multi-Scale Framework for Human Pose and Action Classification
Full text linkHuman pose classification, along with related tasks such as action recognition, is a crucial area in deep learning due to its wide range of applications in assisting human activities. Despite significant progress, it remains a challenging problem because of high inter-class similarity, dataset noise, and the large variability in human poses. In this paper, we propose a lightweight yet highly effective modular attention-based architecture for human pose classification, built upon a Swin Transformer backbone for robust multi-scale feature extraction. The proposed design integrates the Spatial Attention module, the Context-Aware Channel Attention Module, and a novel Dual Weighted Cross Attention module, enabling effective fusion of spatial and channel-wise cues. Additionally, explainable AI techniques are employed to improve the reliability and interpretability of the model. We train and evaluate our approach on two distinct datasets: Yoga-82 (in both main-class and subclass configurations) and Stanford 40 Actions. Experimental results show that our model outperforms state-of-the-art baselines across accuracy, precision, recall, F1-score, and mean average precision, while maintaining an extremely low parameter count of only 0.79 million. Specifically, our method achieves accuracies of 90.40% and 87.44% for the 6-class and 20-class Yoga-82 configurations, respectively, and 94.28% for the Stanford 40 Actions dataset
Artificial Intelligence and the Crises of Judicial Power: (Not) Cutting the Gordian Knot?
Full text linkCourts across the world are experiencing efficiency and contestation crises. In this context, the question emerges as to the role of automated decision-making and artificial intelligence in addressing and potentially solving these very crises of judicial power. The relationship between digital technologies and judicial power is multi-layered and dynamic. Courts can be both users and regulators of technologies. As users, courts can rely on automated decision-making and artificial intelligence to perform their activities. When algorithmic automation is incorporated in courts’ systems, the guardians of the law are exposed to the ordering power of technology, which, in turn, shapes judicial power. But courts also find themselves increasingly involved in solving legal questions on the use of digital technologies. In so doing, judges regulate algorithms by way judicial interpretation. The chapter illustrates that automated decision-making and artificial intelligence do not offer complete solutions to the crises of judicial power. While these technologies certainly have the potential to solve these crises, they fail to do so because of the complexity of judicial power, requiring high standards of meaningful participatory governance and contestability, and the societal forces that underpin judicial authority and legitimacy
Enlightening: Adorno’s and Kant’s contrasting answers to the question “How to write about the Enlightenment?”
Full text linkThe content of Kant’s Enlightenment text has received much critical reception, but very stance Kant takes as its author has been largely ignored. Similarly, there has been much critical discussion of Horkheimer and Adorno’s “Dialectic of Enlightenment” in terms of the theses they (purportedly) endorse, while their authorial voice has mostly received either no attention or been criticised as problematic rhetoric. In this paper, I take a different approach to both texts, focusing on the two
respective writerly stances. I suggest that Kant’s one harbours an implicit epistemic authoritarianism, in contrast to the self-therapeutic stance Adorno and Horkheimer’s text exemplifies
Development of an antibody reformatting strategy for use in targeted protein degradation therapies for neurodegenerative diseases
Full text linkNeurodegenerative diseases (NDDs) such as Parkinson’s disease (PD) and amyotrophic lateral sclerosis (ALS) are a leading cause of global mortality and place a great financial and emotional burden on healthcare systems and support networks. Misfolded insoluble aggregates and toxic soluble proteins are central hallmarks of these NDDs. α-Synuclein aggregation drives Parkinson’s disease pathology and is a suitable target for selective protein clearance. Biological proteolysis targeting chimeras (bioPROTACs) aim to eliminate disease-causing intracellular proteins using host cell ubiquitination and degradation pathways. Here, I describe a bioPROTAC comprising the E3 ubiquitin ligase domain of CHIP (carboxy terminus of Hsc70-interacting protein) fused to NbSyn87, a nanobody specific for α-synuclein. Successful target degradation was achieved using the CHIP-NbSyn87 bioPROTAC. In contrast, CHIP-based bioPROTACs targeting SOD1, a ubiquitously expressed protein prone to misfold and aggregate in ALS, failed to degrade its target. This work highlighted key parameters for consideration during BioPROTAC design including target half-life and solubility, recognition domain binding affinity, molecular chaperone activity, and interdomain linker optimisation. A strategy for rational repurposing of conformation-specific antibodies into soluble, functional intrabodies was also developed. Candidate antibodies were reformatted into scFvs and their solubility and specificity tested. A panel of misfolding-specific SOD1 intrabodies and conformation-specific α-synuclein intrabodies are described. Almost any antibody can be reformatted as a highly soluble (>70 %) and intracellularly stable intrabody using the described approach, based on the discovery of a strong negative correlation between net charge and intrabody solubility in cell models
Lorentzian-Constrained Holographic Beamforming Optimization in Multi-user Networks with Dynamic Metasurface Antennas
Full text linkDynamic metasurface antennas (DMAs) are promising alternatives to fully digital (FD) architectures, enabling hybrid beamforming via low-cost reconfigurable metasurfaces. In DMAs, holographic beamforming is achieved through tunable elements by Lorentzian-constrained holography (LCH), significantly reducing the need for radio-frequency (RF) chains and analog circuitry. However, the Lorentzian constraints and limited RF chains introduce a trade-off between reduced system complexity and beamforming performance, especially in dense network scenarios. This paper addresses resource allocation in multi-user multiple-input-single-output (MISO) networks under the Signal-to-Interference-plus-Noise Ratio (SINR) constraints, aiming to minimize total transmit power. We propose a holographic beamforming algorithm based on the Generalized Method of Lorentzian-Constrained Holography (GMLCH), which optimizes DMA weights, yielding flexibility for using various LCH techniques to tackle the aforementioned trade-offs. Building upon GMLCH, we further propose a new algorithm i.e., Adaptive Radius Lorentzian Constrained Holography (ARLCH), which achieves optimization of DMA weights with additional degree of freedom in a greater optimization space, and provides lower transmitted power, while improving scalability for higher number of users. Numerical results show that ARLCH reduces power consumption by over 20% compared to benchmarks, with increasing effectiveness as the number of users grows
New Design of Sparse Zero-Correlation-Zone Sequence Sets for Optimal Channel Estimation in (Generalized) Spatial Modulation Systems
Full text linkWithin the zero-correlation-zone (ZCZ), ZCZ sequence
sets exhibit ideal correlation properties, which is highly
advantageous for both wireless communications and radar sensing applications. Recently, to achieve optimal training for spatial modulation (SM), Pai et al. introduced the concept of sparse ZCZ (SZCZ) sequence sets, where each column contains only one non-zero element. In this paper, we first extend the SZCZ sequence set concept by permitting multiple non-zero elements per column, thereby accommodating training design requirements for generalized SM (GSM) systems. Then, we propose a direct construction of SZCZ sequence sets with parameter
(qn+k, qm+n+k, (q−1)qπ(2)−1+(q−2)qπ(3)−1, (qn, qm+n)) based
on restricted extended Boolean functions. Compared to existing works, the proposed SZCZ sequence sets exhibit a greater ZCZ width and can be applied to both SM and GSM training designs simultaneously. Simulation results show that compared with other training sequences, the channel estimation performance is significantly improved when the proposed SZCZ sequence set is used as the training sequences
Editorial: Narrow and general intelligence: embodied, self-referential social cognition and novelty production in humans, AI and robots
Full text linkRobot Learning System Based on Target Localization and Human Demonstration for Medical Examination
Full text linkRobot-assisted ultrasound scanning has been proven to be an effective solution for medical examination; however, most of the existing work is fully teleoperation or semiautonomous due to dynamic tasks and uncertain environments. In this article, we propose a novel robot learning system for autonomous ultrasound scanning that integrates dynamic target localization with motion generation based on learning from demonstration (LfD). First, a new LfD model, Gaussian process dynamic movement primitives (GPDMP), is introduced to encode human motion skills and generate motion trajectories online. The proposed GPDMP framework supports multiple demonstrations and guarantees convergence toward the target even in the presence of external disturbances, such as subject movement during ultrasound scanning. To further enhance generalization to moving subjects through robust target localization, a vision-based localization method is developed to identify and localize the neck region. Neck positioning across individuals with different body shapes is addressed by an improved AlphaPose network, while interference from other moving subjects in the scene is mitigated using an enhanced DeepSORT-based tracking algorithm. The proposed system is capable of automatically identifying and localizing the human neck, generating approach trajectories for the ultrasound probe, and achieving real-time interaction. A robot-assisted neck ultrasound scanning system is implemented to experimentally demonstrate the effectiveness of the proposed framework