Bias Training Research Articles

Police leaders’ support for civilian oversight and implicit bias training: assessing the influence of perceptions of systemic racism

ABSTRACT This study uses a sample of police leaders from across the United States to explore the extent to which their support for civilian oversight boards and implicit bias training, as two aspects of police reform intended to improve racial injustices, relate to their belief that systemic racism is an issue in criminal justice. Presented in the context of symbolic racism theory, we assess how respondents’ willingness to acknowledge systemic racism in criminal justice relates to more supportive perceptions of civilian oversight boards and implicit bias training. These relationships remain significant even when controlling for other individual and agency-level factors that are known to affect perceptions of systemic racism and support for public policies designed to improve racial justice. This study discusses the need for additional research and consideration of how police officers’ racialized attitudes relate to their support for common reform recommendations.

Strengthening Equity, Diversity, and Inclusion in medical education via an intersectional approach

Abstract Background In the wake of COVID-19, equity, diversity, and inclusion (EDI) policies have been taken up in the health field, yet the actual impact of these initiatives, particularly for multiply marginalized individuals, is debated. Despite progress in areas such as diversity promotion, EDI has been critiqued as superficial and as failing to challenge the drivers of health inequity. This presentation discusses how an intersectionality-informed approach can help harness the transformative potential of EDI. Methods This research is informed by a literature review of EDI initiatives in the health field, and is an offshoot of a 4.5 year international project funded by the Canadian Institutes for Health Research focused on evaluating and strengthening intersectionality-informed policy guidance. Results There is a disconnect between EDI commitments and substantive action on health inequities and an intersectional approach can help bridge this divide. We demonstrate this with a case example which brings an intersectional approach to EDI initiatives in medical education - including bias trainings, courses on the social determinants of health, and the engagement of community groups. Drawing on promising equity promoting work across these areas, we show how intersectionality can harness the potential of EDI by bringing to the fore what has often been missing, including the promotion of health stakeholder reflexivity, attention to interacting systems and structures shaping health, and meaningful engagement with underrepresented communities. Conclusions An intersectionality-informed approach to EDI has the potential to enhance the health equity impacts of initiatives within medical education and the broader health field. Key messages • EDI initiatives in health often have limited impact on health inequities. • An intersectionality-informed approach to EDI can help health stakeholders drive transformative action on health inequities.

Inclusive Leadership

The aim of this review paper is to explore the crucial role of inclusive leadership in enhancing organisational culture, innovation, and improving performance in the context of digital industrial revolution. The style of inclusive leadership which emphasises on the value of equity, diversity, and inclusion promotes decision-making processes that involve different perspectives. Key practices are highlighted in this review paper such as diverse recruitment, bias training, and inclusive decision-making, while examining their impact on organisational success. To illustrate the benefits of inclusive leadership, three case studies from diverse industries are used including higher employee engagement, innovation, and sustainable growth. Also practical strategies are offered for leaders with the aim to foster inclusivity and tackle challenges related to technological disparities and unconscious bias.

Decoding and modifying dynamic attentional bias in gaming disorder.

With the spread of smartphones and computer games, concerns have escalated regarding the rising prevalence of gaming disorder. Patients often display attentional biases, unconsciously turning their attention towards gaming-related stimuli. However, attempts to discover and ameliorate these attentional deficits have yielded inconsistent outcomes, potentially due to the dynamic nature of attentional bias. This study investigated neural mechanisms underlying attentional bias state by combining neuroimaging (functional magnetic resonance imaging -fMRI) with an approach-avoidance task tailored to an individual's gaming preference. We conducted a multivariate pattern analysis of endogenous brain activity in 21 participants with probable gaming disorder. Our analyses revealed that activity patterns in the insula tracked temporal attentional bias states specific to gaming stimuli. A broad network of frontal and parietal regions instead appeared to predict a general temporal attentional bias state. Finally, we conducted a proof-of-concept study for 'just-in-time' attentional bias training through fMRI-decoded neurofeedback of insula activity patterns, named decoded attentional bias training (DecABT). Our preliminary results suggest that DecABT may help to decrease the attractiveness of gaming stimuli via a insula- and precuneus-based neural mechanism. This work provides new evidence for the insula as an endogenous regulator of attentional bias states in gaming disorder and a starting point to develop novel, individualized therapeutic approaches to treat addiction.This article is part of the theme issue 'Neurofeedback: new territories and neurocognitive mechanisms of endogenous neuromodulation'.

Learning Robust Sequential Recommenders through Confident Soft Labels

Sequential recommenders that are trained on implicit feedback are usually learned as a multi-class classification task through softmax-based loss functions on one-hot class labels. However, one-hot training labels are sparse and may lead to biased training and sub-optimal performance. Dense, soft labels have been shown to help improve recommendation performance. However, how to generate high-quality and confident soft labels from noisy sequential interactions between users and items is still an open question. We propose a new learning framework for sequential recommenders, CSRec, which introduces c onfident s oft labels to provide robust guidance when learning from user-item interactions. CSRec contains a teacher module that generates high-quality and confident soft labels and a student module that acts as the target recommender and is trained on the combination of dense, soft labels and sparse, one-hot labels. We propose and compare three approaches to constructing the teacher module: (i) model-level, (ii) data-level, and (iii) training-level. To evaluate the effectiveness and generalization ability of CSRec, we conduct experiments using various state-of-the-art sequential recommendation models as the target student module on four benchmark datasets. Our experimental results demonstrate that CSRec is effective in training better-performing sequential recommenders.

An innovative framework to advance workplace equity

Reliance on traditional Diversity, Equity, and Inclusion approaches, such as implicit bias and anti-harassment trainings, have failed to address systemic inequities embedded within organizations. In response, we first interviewed business leaders/practicing managers and then fielded a nationally representative survey of 1,062 businesses in the United States (in partnership with the Society for Human Resource Management). Based on our findings, we developed an innovative, two-dimensional framework, the Equity of Employment Systems Framework, that addresses root causes of workplace inequity. Rather than focusing solely on an overall measure of equity for an organization, this Framework focuses on 7 systemic components (“Levers for Change”) that measure and impact equity in an organization’s 10 Employment Systems. Levels of equity vary greatly among the 10 Employment Systems within organizations: Recruitment & Hiring, Compensation & Benefits, and Orientation & Onboarding representing the strongest measures of equity and Supervision & Mentoring, Job Structures, and Resources & Supports as the least equitable. To help practicing managers strengthen the equity of their Employment Systems, we outline a 5-step process based on our innovating Equity of Employment Systems framework. Our 5-step process can help practicing managers to advance equity within their organizations, creating more inclusive, agile, and innovative businesses.

Best Practices in the Nursing Care of Patients With Injuries From Violence: An Integrative Review.

The number of patients who have experienced violence is increasing worldwide. These patients have specific psychosocial and forensic needs and can present unique challenges to the health care workers caring for them. To identify best practices for the care of patients with injuries from violence in the emergency department or inpatient setting. The framework for integrative reviews by Whittemore and Knafl was used to conduct a literature search in MEDLINE, CINAHL, and ProQuest Nursing and Allied Health databases. Fourteen qualitative, quantitative, and mixed-methods studies from peer-reviewed journals were included in the review. Themes included identifying patients as survivors of intimate partner violence or child abuse, the need to collect forensic evidence, the need for emotional support of patients and their loved ones, support for health care workers caring for patients with injuries from violence, prevention of further violence, and the challenges of caring for patients with injuries from violence with concurrent substance use disorders. Best practices for these aspects of care are summarized on the basis of literature analysis. Screening procedures are needed to identify survivors of intimate partner violence and child abuse. Forensic evidence preservation policies should be in place. For survivors of sexual assault, a sexual assault nurse examiner should be available. Hospital systems should provide psychosocial resources for patients' and health care workers' mental health, implement violence reduction programs, and provide bias training. More research is needed to determine efficacy of care models and best practice.

Deep-learning and data-resampling: A novel approach to predict cyanobacterial alert levels in a reservoir

The proliferation of harmful algal blooms results in adverse impacts on aquatic ecosystems and public health. Early warning system monitors algal bloom occurrences and provides management strategies for promptly addressing high-concentration algal blooms following their occurrence. In this study, we aimed to develop a proactive prediction model for cyanobacterial alert levels to enable efficient decision-making in management practices. We utilized 11 years of water quality, hydrodynamic, and meteorological data from a reservoir that experiences frequent harmful cyanobacterial blooms in summer. We used these data to construct a deep-learning model, specifically a 1D convolution neural network (1D-CNN) model, to predict cyanobacterial alert levels one week in advance. However, the collected distribution of algal alert levels was imbalanced, leading to the biased training of data-driven models and performance degradation in model predictions. Therefore, an adaptive synthetic sampling method was applied to address the imbalance in the minority class data and improve the predictive performance of the 1D-CNN. The adaptive synthetic sampling method resolved the imbalance in the data during the training phase by incorporating an additional 156 and 196 data points for the caution and warning levels, respectively. The selected optimal 1D-CNN model with a filter size of 5 and comprising 16 filters achieved training and testing prediction accuracies of 97.3% and 85.0%, respectively. During the test phase, the prediction accuracies for each algal alert level (L-0, L-1, and L-2) were 89.9%, 79.2%, and 71.4%, respectively, indicating reasonably consistent predictive results for all three alert levels. Therefore, the use of synthetic data addressed data imbalances and enhanced the predictive performance of the data-driven model. The reliable forecasts produced by the improved model can support the development of management strategies to mitigate harmful algal blooms in reservoirs and can aid in building an early warning system to facilitate effective responses.

Trends in Severe Maternal Morbidity Following an Institutional Team Goal Strategy for Disparity Reduction

BackgroundRacial disparities in maternal pregnancy outcomes, specifically in morbidity and mortality, are persistent in the U.S., and a multifaceted approach to mitigating these disparate outcomes is critical. In 2020, our health system committed to reducing severe maternal morbidity (SMM) in Black patients, employing multiple strategic interventions including implicit bias training, regular reporting of a composite SMM metric stratified by race and ethnicity, standardization of best practices, focused efforts for hemorrhage risk reduction, and system-wide team building. ObjectiveThe goal of this study is to investigate trends in SMM by race across this period of concentrated interventions to improve maternal outcomes overall, and specifically for Black patients. Study DesignThis is a retrospective cohort study evaluating all delivery admissions at an academic, urban, tertiary-care hospital in Philadelphia – one site of a health system encompassing five delivery hospitals – over a three-year period from 2019-2021. Data including patient demographics, clinical features, and outcomes were extracted from the electronic medical record (EMR). Self-reported race was categorized as Black vs. non-Black as documented in the EMR. SMM was defined according to established CDC indicators as well as additional codes identified by Vizient for common sources of SMM including hemorrhage, infection, and embolism. Data were analyzed by year with a multivariable logistic regression model including insurance type and obstetric co-morbidity index (OB-CMI), a weighted scoring system accounting for numerous chronic medical conditions and antepartum pregnancy complications. ResultsIn total, 12,339 deliveries were included, 64.6% (N = 8,012) of which were to Black patients. Median OB-CMI score was higher for Black patients at 3 (inter-quartile range [IQR] 1-5) compared to 2 (IQR 1-4) for non-Black patients, p < 0.01. There was a significant decrease in SMM for the entire cohort over the study period (8.5% in 2019 to 6.5% in 2021, p = 0.001), driven by a decreased rate specifically among Black patients (8.9% in 2019 to 6.6% in 2021, p = 0.005) with a non-significant decrease for non-Black patients (7.8% in 2019 to 6.3% in 2021, p = 0.21). The adjusted model similarly demonstrated decreased risk of SMM over time for Black patients (2020 vs. 2019 adjusted odds ratio [aOR] = 0.81, 95% confidence interval [CI] 0.69-0.96; 2021 vs. 2019 aOR 0.73, 95% CI 0.62-0.86). ConclusionDedicated efforts to improve equity in maternal outcomes over a two-year period (2020-2021) in this hospital serving a Black patient majority were associated with a significant decline in SMM, especially among Black patients. This finding demonstrates the success of a high-level, coordinated, and systematic approach in reducing SMM and associated disparities, and is highly consequential in light of the ongoing major epidemic of racial disparities in obstetric outcomes.

Upholding our PROMISE: Increased representation is not enough to foster belonging in graduate medical education.

Sense of belonging supports academic achievement and encourages career endurance. The purpose of this study was to characterize what individual and institutional factors influence one's sense of belonging by describing the experiences of underrepresented in medicine (UIM) paediatric and internal medicine-paediatric residents in the United States. The authors conducted a national survey of paediatric and internal medicine-paediatric residents. The 23-item anonymous web-based survey was distributed between October 2020 and January 2021 and included questions on socio-demographic characteristics and individual perceptions on sense of belonging, value, common mission or values at a program or institution and respect. The authors used linear mixed models and fitted regression models to examine individual factors and environmental factors that impact sense of belonging, value and respect. Across 29 residency programs, 938 (53%) of 1748 residents completed the survey. One hundred sixty-seven (18%) self-identified as UIM. UIM residents had a lower sense of belonging than non-UIM residents [mean (SD) 3.6 (0.87) vs. 4.0 (0.57)]. Black/AA and Hispanic/Latinx residents had the lowest sense of belonging [3.5 (0.82) and 2.8 (0.93), respectively]. UIM residents demonstrated decreased sense of belonging in programs that lacked bias training and where peers discriminated against them. Sense of belonging was increased in programs where they perceived a sense of support, respect or values alignment. Surprisingly, individual UIM resident sense of belonging was not improved by having more UIM residents in a program. UIM paediatric residents experience a decreased sense of belonging during training. Programs can support sense of belonging for UIM residents by demonstrating respect and support for them and by offering systems for bias training and reporting bias and discrimination. Recruiting a greater number of UIM trainees remains important; however, compositional diversity alone does not improve a sense of belonging for UIM residents. There is continued need for structural/institutional change, including addressing institutional culture and structural racism.

Ethical Considerations and Challenges in the Integration of Artificial Intelligence in Education: A Systematic Review

This systematic review examines those challenges in light of data privacy, algorithmic bias, ethical implications, technological hurdles, and acceptance of AI by educators and students. First, data privacy should be a primary concern, as AI systems require extensive data, bringing up the potential for breach and misuse. Secondly, there must be a robust mechanism concerning data protection and against the application of GDPR. Another critical point is algorithm bias: biased training data sets may lead to discriminative decisions that will increase inequalities in education. It talks about AI's impact on teachers and classroom dynamics because the takeover of responsibilities may lower the intensity of necessary human contact. From a technical perspective, there is so much infrastructure and expertise required that too many educational institutions lack, especially in developing countries. In addition, educators themselves may feel that the change resists and fears job loss and therefore acts as a deterrent to AI integration. The review underscores the imperative for extensive training of teachers to support enabling the integration of AI. It now demands a collaborative effort on the part of all stakeholders to maximize the gains and reduce the drawbacks of AI in educational aspects. Continuous research in, policy-making for, and ethical guidelines on AI are required to benefit all aspects of education equitably and effectively.

The nature and validity of implicit bias training for health care providers and trainees: A systematic review.

The number of health care educational institutions/organizations adopting implicit bias training is growing. Our systematic review of 77 studies (published 1 January 2003 through 21 September 2022) investigated how implicit bias training in health care is designed/delivered and whether gaps in knowledge translation compromised the reliability and validity of the training. The primary training target was race/ethnicity (49.3%); trainings commonly lack specificity on addressing implicit prejudice or stereotyping (67.5%). They involved a combination of hands-on and didactic approaches, lasting an average of 343.15 min, often delivered in a single day (53.2%). Trainings also exhibit translational gaps, diverging from current literature (10 to 67.5%), and lack internal (99.9%), face (93.5%), and external (100%) validity. Implicit bias trainings in health care are characterized by bias in methodological quality and translational gaps, potentially compromising their impacts.

Implicit bias training for police: Evaluating impacts on enforcement disparities.

The purpose of this study was to estimate the behavioral impacts of training police officers in implicit bias awareness and management. Training police in implicit bias reduces racial and ethnic disparities in stops, arrests, summonses, frisks, searches, and/or use of force. A cluster randomized controlled trial using the stepped wedge design was applied to 14,471 officers in the New York City Police Department, with a 1-day training delivered to clusters of police commands between May 2018 and April 2019 and outcomes measured with police records of individual events from April 2018 to May 2019. Police records were supplemented with survey data on 1,973 officers matched to administrative data. For each type of enforcement action, the likelihood that the action involved or was taken against Black or Hispanic suspects, respectively, relative to White suspects was estimated, controlling for potential confounders. Additional analysis allowed for estimating training effects of different magnitudes for Black, Hispanic, and White officers and for officers with greater motivation to act without prejudice or greater concern about discrimination. None of the estimated training effects achieved statistical significance at the .05 level. Isolated and weak evidence of behavioral impacts of the training was detected. Several explanations for the null findings are considered. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Cross-domain fault diagnosis network based on attributes and features transfer with dual classifier under limited and unbalanced datasets

Abstract Deep learning (DL)-based methods have demonstrated significant success in fault diagnosis owing to their robust feature extraction and non-linear fitting capabilities. Meanwhile, their remarkable performance is accompanied by constant operating conditions and sufficient monitoring data. However, in real engineering environments, variable working conditions or limited and unbalanced data are common, which can widen the gap between fault diagnosis methods and real industrial applications. In this paper, we proposed a cross-domain fault diagnosis network based on a dual classifier (CFDNet) with input being limited and unbalanced data to learn attributes and features for unsupervised domain adaptation. We found that the diagnostic performance is commonly bounded by the underlying knowledge, especially feature extraction from original data. Therefore, we designed a new feature encoder with features and relationships, i.e. using a convolutional neural network and graph convolutional network, which improves extraction efficiency while retaining valuable information. Then, we discovered that enforced feature transfer can lead to negative transfer. To mitigate this, we present a feature and attribute transfer framework, which not only achieves features transfer but also enables attributes transfer. Furthermore, it was noted that limited and unbalanced datasets can introduce label bias and lead to biased model training. Hence, we designed dual classifiers to improve the probability of high-confidence final prediction by synthesizing diagnostic results. Comprehensive experiments conducted on three case studies demonstrate the effectiveness and superiority of our method for cross-domain fault diagnosis under limited and unbalanced datasets, which outperforms state-of-the-art methods in this study.

Auditory-motor adaptation: induction of a lateral shift in sound localization after biased immersive virtual reality training

IntroductionSensorimotor adaptation has often been studied in the visual modality through the Prism Adaptation (PA) paradigm. In this paradigm, a lateral shift in visual pointing was found after wearing prismatic goggles. An effect of PA has sometimes been observed on hearing, in favor of a cross-modality recalibration. However, no study has ever shown if a biased auditory-motor adaptation could induce this lateral shift, which appears essential to a better understanding of the mechanisms of auditory adaptation. The present study aimed at inducing an auditory prism-like effect.MethodsSixty healthy young adults underwent a session of active audio-proprioceptive training in immersive virtual reality based on Head Related Transfer Functions (HRTF). This training consisted of a game in which the hand-held controller emitted sounds either at its actual position in a control group or at 10° or 20° to the right of its actual position in two experimental groups. Sound localization was assessed before and after the training.ResultsThe difference between both localization tests was significantly different between the three groups. As expected, the difference was significantly leftward for the group with a 20° deviation compared to the control group. However, this effect is due to a significant rightward deviation in the control group whereas no significant difference between localization tests emerged in the two experimental groups, suggesting that other factors such as fatigue may have cumulated with the training after-effect.DiscussionMore studies are needed to determine which angle of deviation and which number of sessions of this audio-proprioceptive training are required to obtain the best after-effect. Although the coupling of hearing and vision in PA still needs to be studied, adding spatial hearing to PA programs could be a promising way to reinforce after-effects and optimize their benefits.

The Guardians of the Streets: Frontline Law Enforcement Officers

This study explores the intricate realm of frontline law enforcement personnel, the sentinels who patrol the streets and deal with the day-to-day challenges of maintaining community safety. The aim is to investigate the obstacles and prospects they face, with a particular emphasis on mental health, de-escalation strategies, and community involvement. Principal Subjects: The Effects of Prolonged Stress: We will look at how common stress and trauma-related illnesses like depression and PTSD are in front-line law enforcement personnel. De-escalation Techniques and implicit bias the study will examine how well de-escalation training programs reduce the need for force as well as how implicit bias affects officers' judgment when they engage with members of the public. Establishing trust through community Policing. This section will look at ways to help frontline police and the communities they serve build strong bonds, with an emphasis on underserved areas. Principal Arguments Frontline officers' mental health is greatly impacted by ongoing stress. Officers who receive specialized de- escalation training may be better able to settle disputes amicably and use less force. Officers' interactions with the public can be less affected by unconscious stereotypes when they get implicit bias training. Building trust is a key component of effective community policing tactics that can boost public safety and strengthen ties within the community. Significance by looking into these topics, the research hopes to. Develop efficient stress management methods that will improve officer well-being. Implementing evidence-based de-escalation training will improve public safety. improved community-police ties by implementing programs aimed at fostering trust. Target audience: A wide range of people are intended to read this research, including Policymakers and law enforcement leaders who can put the findings into practice by implementing better policies. mental health specialists who are capable of creating focused interventions to enhance the wellbeing of officers. Community organizations trying to bridge the gap between law enforcement and the public. Additionally, the general public is interested in learning more about the opportunities and difficulties that frontline law enforcement officers encounter.

Investment, integration, and innovation: Fostering diversity, equity, and inclusion across the HEALthy Brain and Child Development (HBCD) Study consortium

The HEALthy Brain and Child Development (HBCD) Study, a multi-site prospective longitudinal cohort study, will examine human brain, cognitive, behavioral, social, and emotional development beginning prenatally and planned through early childhood. From the outset, the HBCD Study integrated diversity, equity, and inclusion (DEI) into its core mission, with key elements like an Associate Director for DEI, a DEI values statement, and establishing a DEI committee to help ensure sustainable progress for the future. The DEI Coordinating Committee supports DEI efforts impacting members of the HBCD consortium, study participants, and the analysis and dissemination of HBCD data. Committee members include representatives from every study site and workgroup which contributes to DEI integration throughout the study. Committee activities include reviewing all measures and protocols, creating accountability metrics, and supporting training opportunities for consortium members. Several successes and lessons have been learned through the Committee’s activities like implementing consortium-wide unconscious bias training and changing the DEI leadership and committee structure. This article presents an overview of HBCD’s DEI components within HBCD and provides examples of collaborative efforts between the DEI Coordinating Committee and other workgroups. The article concludes with plans for future activities and recommendations for other large consortia considering formal DEI structures.

Causality-inspired crop pest recognition based on Decoupled Feature Learning.

Ensuring the efficient recognition and management of crop pests is crucial for maintaining the balance in global agricultural ecosystems and ecological harmony. Deep learning-based methods have shown promise in crop pest recognition. However, prevailing methods often fail to address a critical issue: biased pest training dataset distribution stemming from the tendency to collect images primarily in certain environmental contexts, such as paddy fields. This oversight hampers recognition accuracy when encountering pest images dissimilar to training samples, highlighting the need for a novel approach to overcome this limitation. We introduce the Decoupled Feature Learning (DFL) framework, leveraging causal inference techniques to handle training dataset bias. DFL manipulates the training data based on classification confidence to construct different training domains and employs center triplet loss for learning class-core features. The proposed DFL framework significantly boosts existing baseline models, attaining unprecedented recognition accuracies of 95.33%, 92.59%, and 74.86% on the Li, DFSPD, and IP102 datasets, respectively. Extensive testing on three pest datasets using standard baseline models demonstrates the superiority of DFL in pest recognition. The visualization results show that DFL encourages the baseline models to capture the class-core features. The proposed DFL marks a pivotal step in mitigating the issue of data distribution bias, enhancing the reliability of deep learning in agriculture. © 2024 Society of Chemical Industry.

The martensitic structures, exchange bias effect and training effect in Ni-Mn-Ga ribbons

The exchange bias effect is the physical basics of spin valves and magnetic recording devices. A stable exchange bias effect would be suitable for practical application. In this work, Ni50Mn50-xGax (11≤ x ≤16) ribbons were obtained by melt spinning. The grain size and microstructure have been tuned through melt spinning, which changes the exchange interaction in ribbons and thus gets a giant and stable exchange bias effect. As the increase of x, the structure of Ni50Mn50-xGax (11≤ x ≤16) ribbons undergoes a transition from non-modulated martensite to non-modulated martensite &10 layers modulated martensite to 10 layers modulated martensite. The average grain size of all ribbons is less than 10 μm, and the martensitic plate widths are smaller than 200 nm. With the increasing x or the cooling fields, the magnetization of the ribbons has been enhanced. As a result, the competition between ferromagnetic and antiferromagnetic interactions has been changed and thus influencing spin glass behavior. Furthermore, a low degree of attenuation for the exchange bias effect after continuous cycles has been achieved which arises from the stable spin configuration with a larger contribution of frozen spin which has a slow relaxation rate. For example, the exchange bias field of 8.78 kOe in Ni50Mn38Ga12 ribbons can be achieved under the 40 kOe field cooling at 10 K, and the exchange bias field after infinite cycles can be kept at 8.58 kOe, the degree of attenuation is only 3.4 %. This work enriches the research on the exchange bias effect of Heusler alloys and provides the experimental basis and theoretical support for designing a stable exchange bias effect.

A two-stage importance-aware subgraph convolutional network based on multi-source sensors for cross-domain fault diagnosis

Graph convolutional networks (GCNs) as the emerging neural networks have shown great success in Prognostics and Health Management because they can not only extract node features but can also mine relationship between nodes in the graph data. However, the most existing GCNs-based methods are still limited by graph quality, variable working conditions, and limited data, making them difficult to obtain remarkable performance. Therefore, it is proposed in this paper a two stage importance-aware subgraph convolutional network based on multi-source sensors named I2SGCN to address the above-mentioned limitations. In the real-world scenarios, it is found that the diagnostic performance of the most existing GCNs is commonly bounded by the graph quality because it is hard to get high quality through a single sensor. Therefore, we leveraged multi-source sensors to construct graphs that contain more fault-based information of mechanical equipment. Then, we discovered that unsupervised domain adaptation (UDA) methods only use single stage to achieve cross-domain fault diagnosis and ignore more refined feature extraction, which can make the representations contained in the features inadequate. Hence, it is proposed the two-stage fault diagnosis in the whole framework to achieve UDA. In the first stage, the multiple-instance learning is adopted to obtain the importance factor of each sensor towards preliminary fault diagnosis. In the second stage, it is proposed I2SGCN to achieve refined cross-domain fault diagnosis. Moreover, we observed that deficient and limited data may cause label bias and biased training, leading to reduced generalization capacity of the proposed method. Therefore, we constructed the feature-based graph and importance-based graph to jointly mine more effective relationship and then presented a subgraph learning strategy, which not only enriches sufficient and complementary features but also regularizes the training. Comprehensive experiments conducted on four case studies demonstrate the effectiveness and superiority of the proposed method for cross-domain fault diagnosis, which outperforms the state-of-the art methods.