Browse Articles

Feature-indistinguishable machine unlearning via negative-hot label encoding and class weight masking

Abstract With the growing importance of data privacy and regulatory compliance, machine unlearning has become a critical requirement in deep learning. However, existing approaches often require access to the original training data, incur substantial computational costs, or compromise performance on retained data. To address these limitations, we propose a novel unlearning framework that integrates label encoding fine-tuning with class weight masking, enabling efficient and selective forgetting of specific classes. In particular, we introduce Negative-Hot Label Encoding (NHLE), which suppresses the discriminability of target classes in the feature space, thereby weakening their representations. Our method requires only a small number of samples from the forgotten classes for iterative fine-tuning. Extensive experiments on multiple visual datasets show that the proposed framework achieves near-zero classification accuracy on forgotten data, while reducing accuracy on retained data by no more than 0.035.

Discovery of the most compact 3+1-type quadruple star system TIC 120362137

Abstract Hierarchical multiple stellar systems with short outer periods comprise an important subgroup of multiple star systems. In this paper we present the discovery and spectro-photodynamical analysis of the most compact known 3+1 quadruple stellar system, TIC 120362137. Through investigations of the observations made with the TESS satellite and ground-based follow up measurements, we find that the system consists of an eclipsing binary with a few-day-period that in turn eclipses, and is eclipsed by, a third star on a P mid  = 51.3 d orbit. This inner subsystem, which contains three stars that are more massive and hotter than the Sun, is more spatially compact than Mercury’s orbit around our Sun, and is orbited by a fourth Sun-like star with a period P out  = 1046 d. We detect the spectral lines of all four stars, making this system the most thoroughly studied 3+1 type quadruple stellar system. The future evolution of TIC 120362137 is also modeled, and we conclude that this entire system will likely end up as a pair of white dwarfs.

A dynamic risk prediction framework of lane-changing behavior based on driving intention recognition on icy and snowy surfaces

Canonical BAF chromatin remodeling complex specifies stem cell fate via cell-type-specific co-factor recruitment

Synergistic effect of rubber powder and nano-silica on pore structure and frost resistance of concrete

Declining grassland canopy height in China under asymmetric biomass allocation

Abstract Grassland canopy height is one of the most important traits for determining plant diversity and community structure, directly affecting the resource use efficiency of livestock in grassland ecosystems. However, broad-scale changes in grassland canopy height are seldom reported due to the complex effects of species aggregation on both interspecific and intraspecific structures. Here, we decouple grassland aboveground biomass into vertical and horizontal allocations, thereby offering a pathway to mirror changes in grassland canopy height. Grassland aboveground biomass is estimated using a machine learning algorithm by fusing climatic factors, satellite-driving metrics, and 8-year consecutive ground-truth surveys; the horizontal allocation of grassland aboveground biomass is derived from optimized linear spectral mixture analysis. We find that changes in horizontal biomass allocation primarily accounted for increases in Chinese grassland aboveground biomass from 2001 to 2022, resulting in a significant decline in grassland canopy height. The decline in grassland canopy height is shaped by reduced radiation and, more importantly, by the combined effects of warming and grazing, while also being related to variations in plant diversity. The dwarfing grassland community with declining canopy height may increase the impact of livestock disturbances, thus diminishing the resistance of grassland ecosystems to climate fluctuations.

Enhancing variable frequency drive efficiency using fractional hybrid Particle Swarm Optimization and comprehensive thermal management

Universal work extraction in quantum thermodynamics

Expertise shapes the kinematic and electromyographic characteristics of on-ice side-cutting in elite versus beginner ice hockey players

GPR146 in adipose tissue drives adipose-liver crosstalk and promotes hepatic steatosis in mice

Abstract The limited therapeutic options for metabolic dysfunction-associated steatotic liver disease (MASLD) underscore the need for deeper mechanistic insight and new treatment strategies. Here, we identify the orphan G protein-coupled receptor GPR146 as a regulator of hepatic steatosis through adipose–liver crosstalk. Human genetic analyses link the GPR146 locus to circulating markers of liver injury and inflammation. In mice, both constitutive and acute GPR146 depletion protect against diet-induced obesity and hepatic steatosis. Notably, adipose-specific, but not liver-specific, GPR146 deletion reduces hepatic lipid accumulation by limiting free fatty acid (FFA) influx. Mechanistically, GPR146 promotes adipogenesis in preadipocytes via Gαq-PKC-AKT signaling, increasing lipid storage capacity, and enhances lipolysis in mature adipocytes through ERK activation, elevating circulating FFA. Together, these coordinated actions increase FFA delivery to the liver, promoting triglyceride accumulation. Our findings establish GPR146 as a pleiotropic regulator of adipose tissue biology and a potential therapeutic target for MASLD.

Simulation of phased alerting of community first responders for cardiac arrest

Background Community First Responders (CFRs) are commonly used for out-of-hospital cardiac arrests, and advanced systems send so-called phased alerts: notifications with built-in time delays. The policy that defines these delays affects both response times and volunteer fatigue. Methods We compare alert policies by Monte Carlo Simulation, estimating patient survival, coverage, number of alerts and redundant CFR arrivals. In the simulation, acceptance probabilities and response delays are bootstrapped from 29,307 rows of historical data covering all GoodSAM alerts in New Zealand between 1-12-2017 and 30-11-2020. We simulate distances between the patient and CFRs by assuming that CFRs are located uniformly at random in a 1-km circle around the patient, for different CFR densities. Our simulated CFRs travel with a distance-dependent speed that was estimated by linear regression on observed speeds among those responders in the above-mentioned data set that eventually reached the patient. Results The alerting policy has a large impact on the four metrics above, and the best choice depends on volunteer density. For each volunteer density, we are able to identify a policy that improves GoodSAM New Zealand’s current policy on all four metrics. For example, when there are 30 volunteers within 1 km from the patient, sending out alerts to 7 volunteers and replacing each volunteer that rejects by a new one, is expected to save 10 additional lives per year compared to the current policy, without increasing volunteer fatigue. Our results also shed light on polices that would improve one metric while worsening another, for example, when there are 10 volunteers within 1 km from the patient, dispatching them all immediately increases our survival estimate by 11% compared to the current policy, with the downside of also increasing the redundant arrivals by 137%. Conclusions Monte Carlo simulation can help CFR system managers identify a good policy before implementing it in practice. We recommend balancing survival and volunteer fatigue, aiming to ultimately further improve a CFR system’s effectiveness.

Proof-of-concept thermoelectrically cooled sow-cooling mat

Climate’s influence on topography encoded in stream network topology and geometry

Abstract Stream networks express how Earth’s hydrologic cycle is embedded within its three-dimensional topography. In a top-down view, a stream network’s morphology is often described by its topological connectivity and branching geometry. Although these two characteristics are naturally connected, they have mostly been studied independently, leaving their co-evolution poorly understood. Here, we analyze the topology and geometry of 16,322 5 th -order real-world stream networks across the contiguous United States, showing how they are shaped by climate and the evolution of Earth’s topography. We find that ~73% of these networks show topological self-similarity in their branching patterns and that small tributaries join larger streams at systematically wider angles. Our analysis further reveals that correlations between climate and network topology observed in other studies are mainly mediated through the climate-dependence of networks’ geometric and topographic properties, such as their junction angles and channel slope ratios of merging tributaries. These findings demonstrate the co-evolution of network geometry, topography, and topology under the influence of landscape evolution driven by climatic forcing.

The use of immersive virtual reality technology in dementia care education for nursing students: A scoping review protocol

Objective This scoping review aims to identify and map the current available evidence regarding the use of immersive virtual reality technology in dementia care education for nursing students. Introduction With the growing population living with dementia, there is an urgent need to prepare nurses for quality dementia care through providing relevant dementia care education for nursing students. Virtual reality is an emerging education medium for nursing education and dementia care education for nursing students. However, little evidence focuses on mapping how immersive virtual reality technology has been used in dementia care education for nursing students. With the increasing development and use of this technology, it is crucial to explore current evidence regarding the content and learning outcomes, design and development, and implementation of immersive virtual reality technology in dementia care education for nursing students. Inclusion criteria Studies are eligible if they describe the use of immersive virtual reality technology in dementia education. The dementia care education must be designed for or targeted at nursing students or healthcare professional students (including nursing students) within academic settings, such as nursing schools, universities, and colleges. Methods The proposed scoping review will be conducted following the Joanna Briggs Institute methodology for scoping reviews and the Population, Concept and Context (PCC) framework. The review will search CINAHL, MEDLINE, PsycINFO, Scopus, Web of Science and ProQuest Dissertations and Theses Global. Grey literature search will be conducted via Google Scholar. Extracted data on authors, country, study design, setting, and findings relevant to the review objective will be presented in a data extraction table, accompanied by a quantitative summary of studies’ characteristics and narrative summary. The PRISMAS-ScR figure will detail the selection process. Review registration number This study has been registered with the Open Science Forum (OSF).

Conservation tillage and sprinkler irrigation for sustainable water management and enhanced crop yields in maize and field pea cropping system

Moisture-responsive crystallization strategy for efficient CsPbI3 solar cells fabricated under high-humidity conditions

Phthalates exposure and serum uric acid level in patients with Crohn’s disease: A cross-sectional study

Objective Crohn’s disease (CD) is closely associated with disorders of uric acid metabolism. Our previous research found an association between phthalate exposure and oxidative stress in CD, suggesting a potential role for phthalates in metabolic disorders. Therefore, this study aims to examine their influence on uric acid metabolism in patients with CD. Methods We designed a cross-sectional study involving 117 patients with CD. Ten urinary phthalates metabolites (mPAEs) were detected by gas chromatography-tandem mass spectrometry, and the serum uric acid (SUA) levels were tested. Correlation analysis and Bayesian kernel machine regression (BKMR) models were applied separately to evaluate the associations. Results The prevalence of hyperuricemia was 12.8% (15/117) in CD patients. None of them were obese or had abnormal renal function. In males, we identified significant positive associations between SUA and eight mPAEs (MMP, MIBP, MBP, MBzP, MOP, MEOHP, MEHHP, & MECPP). However, no positive associations between mPAEs and SUA were found in females. After BKMR analysis and multivariate adjustments, we found that the average SUA (μmol/L) increased by 1.36-fold, and the odds ratio for hyperuricemia increased by 1.25-fold, when overall phthalates exposure increased from 25% to 75% in male CD patients. This suggests a potential link between phthalates exposure and uric acid metabolism in male patients with CD. Furthermore, oxidative stress mediated approximately 5% of the association, indicating it is a partial, but not primary, mechanism in this process. Conclusions Phthalates exposure positively correlated with SUA in male CD patients. Effective PAE exposure control in patients with CD may reduce the risk of hyperuricemia.

A multiplex allele-specific polymerase chain reaction assay for rapid and affordable detection of APOL1 risk variants

Cost-effective strategies can reduce water and energy requirements in China’s wastewater treatment by 2035

Tobacco product cessation and prenatal care utilization: A Pregnancy Risk Assessment Monitoring System phase 8 study of conventional cigarette, e-cigarette, and dual-use associated behaviors

Background Pregnancy is a critical window for tobacco cessation; socio-demographic correlates and prenatal care (PNC) utilization behaviors associated with quitting e-cigarette and dual use with conventional cigarettes remain understudied. Objective To examine the influence of socio-demographic characteristics and PNC utilization in the cessation of cigarette, e-cigarette, and dual-use during pregnancy, using Pregnancy Risk Assessment Monitoring System (2016–2022) data. Methods We analyzed data from 223,793 respondents (weighted count = 11,475,844) with singleton births who reported cigarette and/or e-cigarette use during the three months before and the last three months of pregnancy. Socio-demographic and PNC cessation correlates for cigarette, e-cigarette, and dual-use were examined versus continuation using logistic regression analysis. The Adequacy of Prenatal Care Utilization (APNCU) Index assessed PNC. Associations were expressed as adjusted odds ratios with 95% confidence intervals [AOR (95%CIs)]. Results Tobacco product use declined from 16.9% pre-pregnancy to 7.5% by late pregnancy. Pregnancy-associated cessation rates were 53.7% for cigarettes, 81.0% for e-cigarettes, and 48.1% for dual-use. Primiparity was associated with higher odds of quitting across all groups: cigarettes [1.8 (1.6‒2.0)], e-cigarettes [1.6 (1.2‒2.1)], and dual-use [2.3 (95% CI: 1.7‒3.1)]. Black race and Hispanic ethnicity were positively associated with cessation of cigarettes and dual-use, while Black race was also associated with higher odds of EC cessation. A higher smoking frequency was associated with reduced cessation odds of cigarette [0.2 (0.2‒0.3)] and dual-use [0.3 (0.2‒0.4)], while a higher vaping frequency was associated with reduced e-cigarette cessation [0.3 (0.2‒0.5)]. Inadequate APNCU Index was associated with lower odds of quitting cigarettes [0.6 (0.5‒0.7)] and dual-use [0.6 (0.4‒0.9)], but showed no significant association with e-cigarette cessation [0.9 (0.6‒1.5)]. Conclusions Distinct socio-demographic and PNC factors influence cessation patterns by product type. Findings underscore potential opportunities to integrate PNC with targeted cessation support, particularly for high-risk groups amid rising e-cigarette and dual use during pregnancy.