Robust domain adaptation with noisy and shifted label distribution
Peer-Reviewed Publication
Updates every hour. Last Updated: 20-Aug-2025 21:11 ET (21-Aug-2025 01:11 GMT/UTC)
Research team proposed RSLR, a robust self-training approach with label refinement for unsupervised domain adaptation under label noise. It uses LNet for pseudo-labeling and TNet for target-specific training, achieving successful performance on benchmark datasets.
Research team designed PBCounter, a weighted model counting solver for pseudo-Boolean formulas. It uses variable elimination and dynamic programming with ADDs, outperforming state-of-the-art CNF-based solvers in experiments.
The protective effects of Lepidium draba L. on cyclophosphamide-induced hepatotoxicity and nephrotoxicity are explored in rats.
An artificial intelligence (AI) model trained to detect blocked coronary arteries based on electrocardiogram (ECG) readings performed better than expert clinicians and was on par with troponin T testing, according to research presented at the American College of Cardiology’s Annual Scientific Session (ACC.25). The findings suggest that the freely available open-source AI model could help physicians more quickly identify patients who require urgent treatment for a heart attack.
Recently, the research team led by Professor Heping Zhang at the Key Laboratory of Dairy Biotechnology and Engineering, Ministry of Education, Inner Mongolia Agricultural University, published a groundbreaking study in the prestigious journal Science Bulletin. Titled "A Genomic Compendium of Cultivated Food-Derived Lactic Acid Bacteria Unveils Their Contributions to Human Health", the study details the development of the comprehensive genomic dataset of food-derived lactic acid bacteria (FLAB). By systematically revealing their diversity and potential applications in food fermentation and human health, this research provides valuable scientific insights and foundational support for future studies and practical applications of LAB.
In a paper published in Science Bulletin, an international team of scientists examined the extent changes in China’s wetlands from 1980 to 2020 and highlighted the hidden loss and restoration effects based on China_Wetlands product. This consistent dataset (i.e., China_Wetlands) has been created by applying the HOHC method to over 53,000 Landsat images acquired in six time periods from 1980 to 2020.
In a paper published in National Science Review, scientists present evidences that tidal forces have a significant impact on temporal variations in seismic velocity within fault fracture zones.