3789 research outputs found
Sort by
Perceptions of women disclosing intermittent self-catheterization: the role of empathy and disgust
No full textThe behavioural immune system evolved to detect and avoid pathogens, triggering disgust and
aversive cognitive and behavioural responses. Non-contagious cues resembling infection (e.g.,
bodily fluids) are often misperceived as pathogenic, potentially contributing to stigma
experienced by individuals with disabilities. While empathy is associated with more positive
attitudes, little is known about response to disability disclosure in friendship contexts. This
experiment is the first to examine the role of the behavioural immune system, empathy, and
disability disclosure in perceptions of intermittent self-catheterization (ISC). Participants read a
vignette about a female who discloses (or does not disclose) an invisible disability (e.g., ISC) and
completed measures of their affective, cognitive, and behavioural perceptions; desire for her
friendship; disgust; and empathy. Behavioural immune system activation did not significantly
predict perceptions of ISC alone. However, empathy was a significant predictor, and the
behavioural immune system did explain additional variance in negative perceptions when
empathy was statistically controlled. Contrary to the hypothesis, the woman disclosing ISC was
perceived more positively and less negatively than the control woman when social desirability
was controlled, with no differences in perceptions between the ISC and colostomy conditions.
Group effects were strongest for cognitions. Level of disclosure (low vs. high) did not
significantly influence perceptions of the woman disclosing ISC. These findings suggest
disability disclosure in the context of ISC use is not associated with stigma. Further, findings
suggest that strategies focused on fostering empathy and targeting cognitions may help promote
more inclusive attitudes toward individuals with disabilities
Prevalence of pain and receipt of pain management in persons receiving inpatient psychiatric services in Ontario
Full text linkBackground: Pain significantly impacts health and quality of life but is often underreported and poorly managed among persons with mental illness.
Objective: The goal was to examine the prevalence of pain and receipt of pain management among individuals who have received inpatient psychiatric services in Ontario, and to identify characteristics associated with receipt of pain management.
Methods: Anonymized population-level data were analyzed from the Ontario Mental Health Reporting System, based on the interRAI Mental Health assessment. Pain prevalence was based on the interRAI Pain Scale with a score that is ≥1. Bivariate and multivariate analyses considered personal (age, sex), social (partner status, residence admitted from, financial trade off in previous month, reported trauma), functional (cognitive performance), and clinical (previous psychiatric admissions, highest number of alcoholic drinks in single sitting, substance use in last month, health instability, psychiatric diagnoses, intellectual or developmental disability, and multimorbidity) to identify factors linked to receiving pain management.
Results: Approximately 22% (n=69,529) experienced pain, but only 49% (n=34,470) of those in pain received pain management. At the multivariate level, several factors were significantly associated with increased odds of receiving pain management, including age 25–44 years (OR=1.164, 95% CI=1.096-1.236), 45-64 years (OR=1.219, 95% CI=1.147-1.295), financial trade-off (OR=1.185, 95% CI=1.113-1.260), use of opiates in the previous week (OR=1.751, 95% CI=1.661-1.845), self-reported trauma (OR=1.477, 95% CI=1.414-1.542), and a diagnosis of substance use disorder (OR=1.247, 95% CI=1.151-1.352). Conversely, residence prior to admission (hospital: OR=0.812, 95% CI=0.782-0.843); homeless: OR=0.816, 95% CI=0.739-0.901); correctional facility: OR=0.721, 95% CI=0.609-0.853), substance use in the last week (inhalants: OR=0.667, 95% CI=0.534-0.833; stimulants: OR=0.919, 95% CI=0.850-0.994), psychiatric diagnoses (psychotic disorders: OR=0.701, 95% CI=0.644-0.764; bipolar disorders: OR=0.770, 95% CI=0.701-0.845; depressive disorders: OR=0.837, 95% CI=0.773-0.906; personality disorders: OR=0.765, 95% CI=0.689-0.849; medication-induced disorders: OR=0.604, 95% CI=0.411-0.887; intellectual or developmental disability: OR=0.848, 95% CI=0.775-0.928), and health instability (OR=0.733, 95% CI=0.703-0.763) were significantly associated with lower odds of pain management.
Conclusion: Less than half of those in pain received pain management. Current findings highlight personal, social, functional, clinical and service use factors are associated receipt of pain management. Future research should prioritize a holistic approach to care in inpatient psychiatry that includes appropriately addressing pain
Zero-shot pruning of transformer language models using non-dominated sorting genetic algorithm
No full textLarge Language Models (LLMs) are advanced neural networks trained on massive text corpora to understand and generate human language. While they have grown significantly in power and capability, their extensive parameter counts result in high computational costs. Current pruning techniques typically apply uniform sparsity across all layers of LLMs. However, not all layers contribute equally to the model's performance.
Therefore, we propose a non-uniform sparsity mapping algorithm that assigns varying levels of sparsity to each layer according to its impact on the model's performance. To identify the optimal effective allocation schemes, we create a search space comprising a population of candidates for sparsity mapping in the LLM. We leverage an evolutionary algorithm to conduct crossover and mutation on the top-performing candidates within this population, guided by performance evaluations. To determine the optimal sparsity mapping, we employ the Non-dominated Sorting Genetic Algorithm II (NSGA-II), which provides us with a set of optimal solutions that balance pruning ratio and performance trade-offs.
We implement an unstructured pruning approach to maximize sparsity. The pruning process is conducted on a sorted list of weights in each layer, utilizing Hessian approximation (the second-order term of the Taylor series expansion) as the basis for selection.
Furthermore, we employ a novel technique for efficiently measuring the importance scores of LLM layers. In this approach, the LLM is divided into several chunks of layers, and at each iteration, the importance score for the selected chunk is computed. By utilizing a gradient accumulation technique, we collect the scores for mini-batches of input data. We applied our algorithm on GPT2 architecture to demonstrate the applicability of our algorithm for LLMs from millions of parameters to billions of parameters.
We perform comprehensive experiments using the Wikitext and PTB datasets, showing that our method leads to substantial performance improvements on the GPT-2 Medium, Large, and XL models. Remarkably, the GPT-2 model pruned using our algorithm achieves a 15.8% and 3.8% smaller model over state-of-the-art techniques, DistilGPT-2 and ZipLM, respectively, while offering less performance degradation. Notably, our approach requires no retraining or fine-tuning, in contrast to these existing methods, which rely on extensive retraining
Insights into the cryptic mating behaviour of Chaga (Inonotus obliquus) (Hymenochaetaceae)
No full textInonotus obliquus(Ach. ex Pers.) Pilàt is a parasitic white heart rot fungus of circumboreal
distribution belonging to the family Hymenochaetaceae. It is primarily hosted on birch, but can
occur on other hardwoods. Typically, this fungus is recognized by a charcoal-like sterile conk with
a yellowish-brown interior that forms on the host tree. Colloquially, this is known as Chaga and is
hailed for its health benefits, cultural significance, and economic value. As the name suggests, the
sterile conk does not produce any basidiospores. Once in an infection cycle of approximately 10 -
80 years, a spore-producing resupinate and poroid basidioma will form beneath the bark of a
recently dead host. This sporadic occurrence makes them difficult to locate and study. The mating
system of this fungus has been speculated to be amphithallic (primarily pseudohomothallic +
tetrapolar heterothallic).
A clarification of general taxonomy, characteristics of decay, life cycle, medicinal properties,
cultivation, conservation, agaricomyceteous mating systems, and the classical/genomic
characteristics of I. obliquus are provided. Wild strains of I. obliquus were isolated from paper
birch in North-Western Ontario and examined for in vitro growth and fruiting characteristics. A
protocol was developed to form fertile fruit bodies reliably under laboratory conditions. The nuclei
within basidiospores were examined using Hoechst 33342 and fluorescence microscopy
throughout the maturation of basidiomata. It was discovered that the basidiospores of I. obliquus
have highly variable nuclear characteristics, as they can contain between 1 and 6 nuclei in different
ratios at different points in development. Comparisons were made between similar work on
Chinese I. obliquus, which showed similarity in some cases and major differences in others.
Speculation on implications of basidiospore nuclear behaviour are made. As this fungus can
reliably fruit in vitro, it may grant us a greater understanding of mating in other members of the
Hymenochaetaceae as a model organism
Affiliative touch and emotional appraisal
No full textMany traditional theories of emotional appraisal hold that emotional meaning is gained only after a stimulus is relayed through the primary sensory cortices and other association areas (centralized appraisal). However, accumulating evidence from non-visual modalities show that the valence of a stimulus can be discerned the moment contact is made with the basic sensory organs (decentralized appraisal). For instance, affiliative touch (e.g., gentle stroking), signals positive affect associated with social contact and support. Critically, however, it remained unknown whether decentralized valence signals influence centralized emotional appraisal mechanisms, such as those processing emotional content from visual stimuli. This is particularly notable given the visual dominance observed throughout much of our sensory processing. The current work addressed this gap in understanding, specifically focusing on how affiliative touch influences emotional appraisal of visual information. Participants were presented with either affiliative touch, neutral touch, or no touch while viewing static images of varied emotional content and were required to rate either the visual or tactile stimuli for its valence and arousal. Affiliative touch was found to have a general influence on the perceived valence of visual stimuli without modulating arousal. However, affectively salient visual stimuli strongly influenced both valence and arousal of tactile stimuli. These findings support the sensory dominance of vision for understanding and interpreting the world, yet also suggests a supramodal influence of affective information, independent of acquisition modality. This study improves our understanding of how affectively salient stimuli from non-visual modalities influence our emotional appraisal of visual information
Defending against false data injection attacks in power systems: techniques for securing state estimation and distance protection
No full textThe digitalization of power systems—enabled by technologies such as digital substations
and Wide-Area Monitoring, Protection, and Control (WAMPAC) systems—has improved
automation, visibility, and system control. However, this increased reliance on data and
communication networks has also increased the system’s vulnerability to cyber threats.
Among them, False Data Injection Attacks (FDIAs) are particularly concerning due to their
stealth and potential to disrupt core grid functions. Their far-reaching impact highlights
the urgent need for comprehensive vulnerability assessments and robust defense strategies
to protect digitalized power systems.
In response to these challenges, this thesis investigates two high-impact FDIA scenarios:
(1) coordinated, stealthy attacks targeting Phasor Measurement Unit (PMU)-based state
estimation, and (2) falsification of protection signals targeting distance relays. For each
case, the thesis first conducts a detailed vulnerability analysis to assess attack feasibility
and impact. Building on these insights, it then develops defense strategies to enhance the
cyber-physical resilience of modern power systems.
The first part of this thesis evaluates the vulnerability of PMU-based state estimation
to multi-step, stealthy FDIAs, in which adversaries coordinate sequential manipulations of
PMU measurements not only to evade bad data detection but also to amplify the cumulative
impact on system operation. To model this attack process, a vulnerability assessment
framework is proposed based on a Markov Decision Process (MDP) integrated with bilevel
optimization. The MDP, solved using Q-learning, models the attacker’s sequential
decision-making and yields a vulnerability index that enables operators to assess system
impact and identify critical attack stages for targeted defense.
This analysis highlights a key insight: while stealthy FDIAs on state estimation typically
require coordinated manipulation of multiple correlated PMUs—an operationally
complex task—compromising a single Phasor Data Concentrator (PDC), which aggregates
data from these PMUs, allows an attacker to simultaneously alter all associated measurements.
This significantly increases the feasibility and potential impact of the attack. Yet,
most defense strategies remain focused on individual PMUs, overlooking the critical role
of PDCs as centralized aggregation points and high-value attack targets.
To address this overlooked threat, the second part of this thesis proposes a tri-level
defender–attacker–operator optimization framework for redesigning PMU-to-Super PDC
(SPDC) assignments as a defense mechanism against stealthy FDIAs targeting state estimation.
The objective is to minimize vulnerability to such attacks while accounting for communication constraints such as transmission delays. Leveraging Software-Defined Networking
(SDN), the framework enables dynamic reassignment of PMUs to SPDCs without
additional cost, providing system operators with a practical and scalable defense strategy.
To further strengthen data aggregation–based defense strategies, it is crucial to consider
not only the assignment of PMUs to PDCs but also the cyber-layer structure—including
communication paths—as both a source of system vulnerability and a target for defense
strategies. Building on this, the thesis analyzes the often-overlooked role of the cyber layer
in vulnerability to stealthy FDIAs and introduces a Cyber-Physical Risk Metric (CPRM)
that combines both the likelihood and physical impact of attacks. The CPRM quantifies
risk by combining the physical consequences of losing a transmission line with the probability
that such a loss results from a stealthy FDIA. This probability is estimated by
identifying minimal critical PMU sets whose compromise could stealthily overload transmission
lines, using an algorithm that solves multiple bi-level optimization problems. Next,
Bayesian Attack Graphs (BAGs) are developed for each substation and communication link
to model potential access pathways and calculate the probability of compromising the identified
critical PMU sets. The thesis then proposes an optimization-based data aggregation
reconfiguration scheme that leverages SDN to dynamically reconfigure both PMU-to-PDC
assignments and their communication paths, minimizing the risk quantified by the developed
metric and serving as a defense mechanism against stealthy FDIAs.
Finally, this thesis addresses FDIAs targeting distance protection and demonstrates
that falsified measurements can severely compromise fault detection and isolation, thereby
threatening power system security and stability. To defend against such attacks, a cyberresilient
protection scheme is proposed, which activates during cyber threats and temporarily
backs up distance relays to maintain system integrity. The proposed protection
scheme mimics the zone-based fault detection of distance relays but leverages traveling
waves (TWs)—which are the natural signatures of real faults—along with dedicated hardwired
current measurements and a Random Forest (RF) classifier to identify faults in each
zone. The RF classifier is trained on the attenuation patterns of TW frequency components
as they propagate from fault locations to the line terminal. Since attenuation patterns depend
on both frequency and travel distance, the RF classifier can accurately determine
the fault zone by extracting frequency-related features from the first TW using a wavelet
transform and analyzing its attenuation characteristics
Access to palliative care by persons with severe and persistent mental illnesses in Ontario
Full text linkObjectives: This study’s aims were to examine access to palliative care by persons with severe and persistent mental illnesses (SPMI), and determine the factors associated with access.
Methods: This study employed a retrospective study design using health administrative data based on the interRAI home care assessment. The prevalence of access to palliative care (PC) among home care clients with and without SPMI was determined. Univariate and multivariate logistic regression models were fitted to assess the association between access to palliative care and the social-demographic and clinical factors that may influence access to palliative care.
Results: Of the 616,296 home care clients, 155,642 (25.3%) had SPMI and 15,057 (2.5%) accessed PC. Of those who accessed PC, 23.5% (3,536) had SPMI. The association between SPMI and PC access was modified by sex (p-value=0.02) and age (p-value=0.04). Females less than 65 years who had SPMI had 15% lower odds (OR=0.85, CI=0.76, 0.95) of PC access compared to males who were more than 65 years and had no SPMI. Also, females aged 65-74 years with SPMI had 16% lower odds (OR=0.84, CI=0.76, 0.93) of PC access compared to males who were not aged 65-74 years and had no SPMI.
Conclusion: Overall persons with SPMI had lower access to PC compared to those without SPMI, a disparity that demands pragmatic healthcare system policy changes to improve access
Tools from above: evaluating drone-borne aerial remote sensing systems for archaeological site and feature identification
Full text linkArchaeological investigations are rapidly changing due to developing digital technologies. They
affect data collection, processing, interpretation, and analysis, but have spawned new approaches
to archaeological investigations. One aspect of this change includes remotely piloted aircraft
systems (RPAS or Unmanned Aerial Vehicles or UAVs, also commonly known as drones) that
have utility in improving cost-effectiveness of site characterization and feature identification but
may not be appropriate for every archaeological situation. These RPAS are rapidly improving, and
becoming more affordable, powerful, and accessible. When employed with digital data processing
methods, they offer an important tool for investigating natural and cultural landscapes. Compared
with imagery from modern satellite and manned aircraft, low altitude drone data offer advantages
in resolution, accuracy, and flexibility. Important emerging considerations involve the
development of diverse drone-deployed sensors coupled with geomatic analysis, machine learning,
and computer-aided enhancement of detected spatial patterns.
This thesis explores the strengths and weaknesses of data collection and processing via aerial
remote sensing, with particular attention to its utility for archaeological detection and
characterization. It evaluates the efficacy and cost-effectiveness of unmanned aerial vehicles
(UAVs) equipped with various sensors to aid archaeological investigation and site analysis.
Further, a variety of data formats were integrated using geographic information systems (GIS).
Information deriving from each of the remote sensing sensors used in this thesis demonstrated
interpretive value. Efforts at validation of non-invasive archaeological interpretation involve direct
visual confirmation of feature anomalies and positive spatial correlation of features of interest
using optical remote sensing, legacy data, and georeferenced imagery.
This study represents the first systematic evaluation of UAVs and sensor technologies for
archaeological use in the Canadian Prairies. The research addressed three key questions through
examples using both consumer and professional-grade UAVs:
1. Can consumer and professional-grade UAVs provide more comprehensive tools for
archaeological site characterization?
2. Can these UAVs help overcome the physical and financial challenges associated with
archaeological fieldwork?
3. What technical and regulatory obstacles hinder the routine integration of consumer and
professional-grade UAVs in archaeological investigations?
These questions are addressed with aerial data from three different archaeological site types from
Manitoba, Canada: the pre-contact Lockport site (case study 1); the fur trade posts at Fort Ellice I
and Fort Ellice II (case studies 2 and 3); and an undisclosed modern/historic cemetery (case study
4)
Childfree stigma
Full text linkThe childfree lifestyle has been gaining increasing mainstream and academic attention, as more
people choose not to have children. Although there are rising numbers of childfree individuals,
stigma remains abundant. The purpose of this set of exploratory studies was to examine
experiences of stigma and self-stigma in childfree people. In Study 1, we examined experiences
of stigma and self-stigma in childfree people using qualitative thematic analysis. Community
members (N = 222) were recruited to complete an electronic survey. Results revealed that most
participants experienced childfree related stigma and self-stigma, with multiple factors
contributing to its development and negative effects. Participant responses also supported the
development of a quantitative scale to measure childfree self-stigma, which was subsequently
created for Study 2. In Study 2, we quantitatively examined stigma and self-stigma in childfree
people, primarily by conducting t-tests, z-tests, and bivariate correlations suited to the
exploratory nature of the data. Childfree community members (N = 440), as well as childfree
university students (n = 125) and non-childfree university students (n = 512) were recruited.
Questionnaires in Study 2 addressed self-stigma, quality of life, trust in healthcare, autonomy,
and personality. The self-stigma scale performed well psychometrically. Main findings further
revealed that self-stigma was negatively correlated with quality of life and autonomy in the
community sample. Further, individuals who had experienced childfree related stigma in the
healthcare system reported reduced trust in healthcare overall. Primary strengths of the project
include the large samples and two-pronged approach to examining the constructs, while
limitations included the cross-sectional and correlational design of the research. This work
supports self-stigma theory and also highlights the ongoing stigma that childfree people face and
the significant challenges and consequences this poses
Enhancing semantic segmentation: architectural innovations and strategies for label-efficient learning
Full text linkSemantic segmentation is a fundamental component of modern computer vision applications. Although
supervised learning models have achieved state-of-the-art performance in this domain, they
rely heavily on large volumes of labeled data, which is an expensive and time-consuming requirement.
Thus, this research aims to develop enhanced supervised semantic segmentation models that
balance accuracy and data efficiency for visual perception tasks in autonomous driving environments.
To achieve this, the thesis is organized into two distinct phases. The first phase investigates
a dual-network architecture, in which an auxiliary boundary detection network is incorporated into
the primary segmentation framework to mitigate pixelation artifacts at object boundaries in multiclass
segmentation of complex scenes. The experimental findings demonstrate the importance of
designing unified segmentation models that take advantage of architectural enhancements capable
of extracting richer feature representations for improved performance. The second phase leverages
insights from the previous stage and focuses on the development of an efficient deep learning
model with attention mechanisms and multi-scale feature refinement. The proposed method introduces
a novel depth-wise, point-wise feature pyramid module that extracts information-rich
spatio-semantic context from early and deep feature representations, improving model efficacy.
Exhaustive experimental studies conducted on widely used benchmark datasets validate the effectiveness
of the proposed models, which achieve competitive performance while offering improved
computational efficiency relative to baseline approaches. The findings highlight that strategically
balancing resource utilization with architectural innovation can yield strong performance while
minimizing annotation demands and environmental impact. This research sets a valuable precedent
for building competitive, resource-aware vision systems suited to constrained application settings