PRESS-NEWS.org - Press Release Distribution
PRESS RELEASES DISTRIBUTION

D2-GCN: a graph convolutional network with dynamic disentanglement for node classification

D2-GCN: a graph convolutional network with dynamic disentanglement for node classification
2025-02-12
(Press-News.org) Classic Graph Convolutional Networks (GCNs) often learn node representation holistically, which would ignore the distinct impacts from different neighbors when aggregating their features to update a node's representation. Disentangled GCNs have been proposed to divide each node's representation into several feature channels. However, current disentangling methods do not try to figure out how many inherent factors the model should assign to help extract the best representation of each node.
To solve the problems, a research team led by Chuliang WENG published their new research on 15 Jan 2025 in Frontiers of Computer Science co-published by Higher Education Press and Springer Nature.
The team proposed a novel disentangled graph convolutional network named D2-GCN to dynamically adjust the number of each node’s disentangled feature channels during training, which helps obtain adaptive node representations on different datasets.
In the research, they design a two-level disentangling mechanism that integrates the epoch-level and the layer-level disentanglement during training. With this mechanism, D2-GCN could capture nuanced changes in node representations on graphs of varying topological complexities. They further leverage the expectation-maximization (EM) algorithm to demonstrate the convergence of the proposed dynamic disentanglement. Finally, they define an information entropy based evaluation metric to portray the convergence speed of the disentangling process.
Experiments show that their model outperforms all the baselines in both single- and multi-label node classification tasks in terms of test accuracies. Visualization results also indicate that D2-GCN could display clearer classification boundaries and higher intra-class similarity than the baseline disentangled methods.
Future work can combine the subgraph theory to improve their model performance in presenting more accurate local-global disentanglement and grasping long-distance dependencies in a graph .
DOI: 10.1007/s11704-023-3339-7
 

END

[Attachments] See images for this press release:
D2-GCN: a graph convolutional network with dynamic disentanglement for node classification D2-GCN: a graph convolutional network with dynamic disentanglement for node classification 2

ELSE PRESS RELEASES FROM THIS DATE:

Female hoverflies beat males on long-distance migrations

Female hoverflies beat males on long-distance migrations
2025-02-12
Male hoverflies are outflown by females when it comes to long-distance migration, new research shows. Marmalade hoverflies leave northern Europe each autumn to escape the cold winter. The study – by the University of Exeter – compared the number of males and females migrating at a northern point (Denmark) and further south (Spain). At the northern point, 50% of hoverflies were male and 50% were female – but at the southern point about 90% were female, suggesting males are “poor long-distance fliers”. “We carried out a range of tests and found females were better adapted for long-distance ...

Study finds consumer openness to smoke-impacted wines, offering new market opportunities

Study finds consumer openness to smoke-impacted wines, offering new market opportunities
2025-02-12
CORVALLIS, Ore. – Certain groups of consumers appear to be open to drinking smoke-impacted wines, a finding in a new study that could provide market opportunities for winemakers increasingly dealing with the effects of wildfire smoke on grapes. The study by researchers at Oregon State University and in New Zealand found that consumers, particularly those that like smokey flavors in food and beverages, are open to drinking smoke-impacted wines. They also found that the type of information on the label can modulate consumer acceptance. “This ...

Why we need to expand the search for climate-friendly microalgae

Why we need to expand the search for climate-friendly microalgae
2025-02-12
New research has highlighted microalgae’s capacity as a solution in the fight against climate change, but researchers warn that “smart microalgal bioprospecting” is needed to unlock its full potential.   The study highlights the vast, largely unexplored capacity of microalgae to mitigate CO2 emissions while driving sustainable industry.  “Microalgae have remarkable properties that make them an ideal tool for tackling climate change,” said lead author, PhD candidate Joan Labara Tirado from the University of Technology Sydney (UTS).  The review paper, The need for smart microalgal bioprospecting was ...

Fewer forest fires burn in North America today than in the past—and that's a bad thing

Fewer forest fires burn in North America today than in the past—and thats a bad thing
2025-02-12
Fewer wildfires burn in North American forests today than in previous centuries, increasing the risk of more severe wildfires, according to research published this week in Nature Communications. The findings may seem counterintuitive, but frequent low-lying surface fires often maintain balance in forests by reducing fuel sources across large areas.  The new study led by the Cooperative Institute for Research in Environmental Sciences (CIRES) at the University of Colorado Boulder and the U.S. Forest Service’s Rocky Mountain Research Station ...

Older people in England are happier now than before the COVID pandemic, new national study suggests

2025-02-12
Older people have greater general happiness, life satisfaction and sense of purpose than they did before the Covid-19 pandemic. That’s according to a new study which tracked 3,999 over 50s in England for 11 years, published today in the peer-reviewed journal, Aging and Mental Health. Analysing data to understand positive psychological well-being and depression within this group, an expert team from UCL (University College London), funded by the National Institute of Ageing, a consortium of UK government departments coordinated by the National Institute ...

Texas A&M chemist wins NSF CAREER Award

Texas A&M chemist wins NSF CAREER Award
2025-02-11
Dr. Alison Altman, an assistant professor in the Texas A&M Chemistry Department, has received a National Science Foundation (NSF) CAREER Award, which supports early-career faculty in research and education. The award recognizes her work on underexplored elements of the periodic table and her dedication to teaching. For Altman, it’s a launchpad for future discoveries. “It’s an honor to receive this award, as it acknowledges not just my research program but also my teaching efforts,” ...

Micro-nano plastics make other pollutants more dangerous to plants and intestinal cells

2025-02-11
Micro- and nanoscale plastic particles in soil and water can significantly increase how much toxic chemicals plants and human intestinal cells absorb, according to two new studies from Rutgers Health that raise fresh concerns about food safety from plastic pollution. The first study in NanoImpact found that lettuce exposed to both nanoscale plastic particles and common environmental pollutants such as arsenic took up substantially more of the toxic substances than plants exposed to the pollutants, alone confirming the risks of polycontamination of our food chain. A companion study in Microplastics journal showed ...

Study of female genital tract reveals key findings

2025-02-11
Seeking to understand what constitutes a healthy vaginal microbiome, a global research collaboration that includes a Rutgers-New Brunswick scientist has reported a series of findings, including identifying which bacteria thwart vaginal disease and determining that microbiomes vary significantly across human populations. Authors of the study, published in Trends in Microbiology, are part of a Belgium-based initiative called the Isala Sisterhood. Members of the group aim to inspire research on microbiomes worldwide by creating a “reference map” of vaginal microbiota. Launched in 2020 at the University of Antwerp, the project has expanded to include more than 3,000 ...

Pitt Engineering Professor Fang Peng elected to National Academy of Engineering

Pitt Engineering Professor Fang Peng elected to National Academy of Engineering
2025-02-11
The National Academy of Engineering (NAE) today announced that University of Pittsburgh Professor Fang Peng, an internationally acclaimed power electronics researcher, is among the newest cohort elected to the academy. The NAE is recognizing Peng for “contributions to the development of high-powered electronic technologies for advanced power grid and energy conversion.” Peng, the RK Mellon Endowed Chair Professor of Electrical and Computer Engineering and Director of the Energy GRID Institute at Pitt’s Swanson School of Engineering, ...

Short-course radiation therapy effective for endometrial cancer patients

Short-course radiation therapy effective for endometrial cancer patients
2025-02-11
In a randomized clinical trial, researchers from Huntsman Cancer Institute at the University of Utah (the U) have found that short-course, higher dose vaginal brachytherapy for endometrial cancer had similar effectiveness to more frequent, lower dose sessions. Gita Suneja, MD, MS, physician-scientist at Huntsman Cancer Institute and professor of radiation oncology at the U, is the first author of the SAVE trial report—which stands for, Short-Course Adjuvant Vaginal Cuff Brachytherapy in Early Endometrial Cancer Compared with Standard of Care. “There ...

LAST 30 PRESS RELEASES:

New software tool could help better understand childhood cancer

Healthy lifestyle linked to lower diverticulitis risk, irrespective of genetic susceptibility

Women 65+ still at heightened risk of cervical cancer caused by HPV

‘Inflammatory’ diet during pregnancy may raise child’s diabetes type 1 risk

Effective therapies needed to halt rise in eco-anxiety, says psychology professor

Nature-friendly farming boosts biodiversity and yields but may require new subsidies

Against the odds: Endometriosis linked to four times higher pregnancy rates than other causes of infertility, new study reveals

Microplastics discovered in human reproductive fluids, new study reveals

Family ties and firm performance: How cousin marriage traditions shape informal businesses in Africa

Novel flu vaccine adjuvant improves protection against influenza viruses, study finds

Manipulation of light at the nanoscale helps advance biosensing

New mechanism discovered in ovarian cancer peritoneal metastasis: YWHAB restriction drives stemness and chemoresistance

New study links blood metabolites and immune cells to increased risk of urolithiasis

Pyruvate identified as a promising therapeutic agent for ulcerative colitis by targeting cytosolic phospholipase A2

New insights into the clinical impact of IKBKG mutations: Understanding the mechanisms behind rare immunodeficiency syndromes

Displays, imaging and sensing: New blue fluorophore breaks efficiency records in both solids and solutions

Sugar, the hidden thermostat in plants

Personality can explain why some CEOs earn higher salaries

This puzzle game shows kids how they’re smarter than AI

Study suggests remembrances of dead played role in rise of architecture in Andean region

Brain stimulation can boost math learning in people with weaker neural connections

Inhibiting enzyme could halt cell death in Parkinson’s disease, study finds

Neurotechnology reverses biological disadvantage in maths learning

UNDER EMBARGO: Neurotechnology reverses biological disadvantage in maths learning

Scientists target ‘molecular machine’ in the war against antimicrobial resistance

Extending classical CNOP method for deep-learning atmospheric and oceanic forecasting

Aston University research: Parents should encourage structure and independence around food to support children’s healthy eating

Thunderstorms are a major driver of tree death in tropical forests

Danforth Plant Science Center adds two new faculty members

Robotic eyes mimic human vision for superfast response to extreme lighting

[Press-News.org] D2-GCN: a graph convolutional network with dynamic disentanglement for node classification