Researchers summarize recent works on using large sequence models like Transformers for sequential decision-making and reinforcement learning tasks, highlighting their potential and limitations. The study categorizes approaches based on data utilization and proposes avenues for future research to improve these models.
Researchers at Nanjing University of Information Science and Technology have developed a guest pre-intercalation strategy to enhance multivalent ion storage. KMO cathode material exhibits high reversible capacity, excellent cycle stability, and potential application in aqueous Al-ion batteries.
Researchers developed a novel method to optimize pump wavelength, reducing ESA-induced absorption and increasing pump quantum efficiency. This approach achieved higher laser output and slope efficiency compared to conventional pumping schemes.
A new scheme for realizing nonreciprocal interlayer coupling in bilayer topological systems has been developed by researchers at Peking University. The scheme is based on constructing on-site gain/loss in bilayer non-Hermitian topological systems and reveals a relation between two microscopic provenances of the non-Hermiticity.
Researchers discovered that intestinal epithelium-derived exosomes, specifically containing miR-21a-5p and miR-145a-5p, reduce liver lipid deposition in mice. These findings suggest a novel pathway for the intestine to regulate liver lipid homeostasis, providing potential targets for NAFLD prevention and treatment.
Researchers developed a machine learning model to predict and optimize VFA production from riboflavin-mediated sludge fermentation. The study found that XGBoost presented the best prediction performance, with high testing coefficient of determination (R2 of 0.93) and low root mean square error (RMSE of 0.070).
A study investigated the risk assessment and remediation of a retired industrial park in Zhejiang Province, China. The researchers measured soil and groundwater samples to assess the comprehensive risk of organic pollutants, highlighting the importance of accurate classification and grading of risk zones.
Researchers discovered that Y-box binding protein 1 (YBX1) in mitochondria inhibits pyruvate entry, facilitating cancer cell proliferation and metastasis. Downregulation of YBX1 promotes cellular oxygen consumption and reduces lactic acid production, while inhibition of the MPC1/2 complex by YBX1 enhances metastatic capacity.
The DeepDrug framework leverages graph-based deep learning to predict drug-target interactions and optimize drug discovery. It achieves optimal performance across various tasks, including binary-class DDI and DTI classification, multi-class/multi-label DDIs, and regression tasks.
Recent publications in Quantitative Biology assess AI's potential to improve life science research, including applications in medicine and bioinformatics. Researchers propose frameworks for integrating AI in digital investigation, such as Digital Life Systems, which aims to comprehensively integrate AI into system modeling.
Regular exercise training boosts memory and spatial cognition in obese mice, reducing age-related cognitive decline and inflammation. The study highlights the crucial role of physical activity in preventing obesity-induced cognitive dysfunction and neurodegenerative diseases.
The research team discovered that chloride ions cause a dramatic negative impact on blue PeLEDs' operational lifetime due to their low migration energy barrier. This issue necessitates innovative strategies to immobilize chloride ions within blue perovskite emitters.
Researchers led by Prof. Ray-Hua Horng published a groundbreaking study on the performance of thin-film vertical-cavity surface-emitting lasers (VCSELs) on composite metal substrate, introducing innovative methods to enhance thermal management. The breakthrough technique brings substantial improvements in device characteristics and pot...
Scientists create techniques for fine-tuning surface features and hierarchies using ultrafast lasers. They demonstrate the ability to build complex structures, such as fort-like formations, by controlling in-situ deposition during laser ablation.
Researchers have successfully developed a tapered fiber-based system that can deliver high-energy ultrafast lasers with near-diffraction-limited beam quality. The system achieved pulse energies of up to 126 μJ and peak powers of 207 MW, representing the highest reported values from a monolithic fiber laser.
LIBRA surpasses existing tools in increasing cell-type resolution and predictive power across paired data sets, especially with adaptive parameter optimization (aLIBRA). The tool's performance is significantly boosted by aLIBRA, outperforming other methods in chromatin accessibility prediction.
Researchers developed a Zr-modified-bentonite filled polyvinyl chloride membrane for efficient phosphate removal from low-concentration phosphate solutions. The membrane achieved high removal rates of up to 98.5% in SCAW, making it a promising solution for combating eutrophication and alleviating phosphorus shortages.
Researchers developed a two-step hydrothermal method to transform biomass waste into humic acid, increasing efficiency and reducing dependence on non-renewable resources. The study optimized acidic hydrothermal humification conditions and explored the mechanism of action affecting humification ability in alkaline hydrothermal treatment.
A new multi-level three-dimensional quantum wavelet transform theory is proposed to implement the wavelet transform for quantum videos, offering exponential speed-up over classical counterparts. The proposed wavelet transforms have better compression performance for quantum videos than two-dimension quantum wavelet transforms.
A new research proposes an Iterative Android Automated Testing (IAAT) method that automatically records and integrates User Operation Processes (UOPs) from manual testing. The IAAT method shows a significant improvement in test coverage compared to traditional automated testing methods.
A recent study reveals that intestinal MCT1 regulates inflammation and metabolism in a sex-dimorphic pattern, with male mice showing enhanced glucose tolerance and reduced inflammation, while female mice experience exacerbation of diet-induced obesity. The study suggests that gender-specific treatments are needed for metabolic disorders.
A team of researchers proposes a novel representation learning method based on an integrated autoencoder for unsupervised domain adaptation. They introduce a sparse autoencoder to combine inter- and inner-domain features, improving performance and minimizing deviations in different domains.
Researchers propose two novel unsupervised spectral feature selection algorithms to detect informative features in high-dimensional data with small samples. The algorithms use advanced clustering and ranking techniques to guarantee representative and independent features, leading to reliable and trustworthy medical diagnostic systems.
Researchers focus on optimizing F-passivated ZnO electron transport layers to improve PbSe colloidal quantum dot photovoltaics. The work aims to decrease trap density and enhance device performance, paving the way for more efficient solar cells.
Researchers have developed a new single crystal material that significantly enhances optical thermometry with Yb,Ho:GYTO, offering high sensitivity and rapid response. The material enables non-contact temperature measurement in harsh environments such as intracellular, coalmines, and power stations.
Researchers developed a mixed solvent system for stable PbS colloidal quantum dot inks, enabling large-scale blade coating of uniform QD films. This resulted in high-performance infrared solar cells with average and filtered PCEs of 11.14% and 4.28%, respectively.
A new approach called Net Learning uses deep neural networks to approximate the reachability set of Petri nets, avoiding the state space explosion problem. This allows for a probabilistic solution to be obtained without needing to solve the equivalent NP-Hard problems.
Researchers developed a gradient boosting-assisted machine learning model to predict free chlorine residual concentrations in drinking water treatment plants. The model demonstrated accurate predictions from cost-effective monitoring data and identified key influencing parameters.
Researchers in China have developed an effective oil separation technique using solar thermal energy, achieving high oil removal efficiency (>99%) and long-term stability. The new protocol also generates electricity and is environmentally friendly.
Researchers discovered that manganese ions can directly bind to the coat protein complex II (COPII) complex, enhancing its condensation and resulting in a unique bell-shaped regulation on blood lipid levels. This mechanism enables the reversal of atherosclerotic plaques in murine disease models.
Researchers introduce FragDPI, a novel method for predicting drug-protein binding affinity by merging sequence information of drugs and proteins. The model yields commendable outcomes compared to baselines, accurately identifying specific interaction parts of drug-target pairs.
Researchers developed a novel deep learning-based model, circ2CBA, to predict circRNA-RBP binding sites. The model achieved good performance and outperformed other methods, particularly in sub-datasets, due to its ability to capture context-dependent information.
A new method, attribute augmentation-based label integration (AALI), enhances crowdsourced label quality by identifying reliable instances and improving the discriminative ability of the original attribute space. Experimental results demonstrate AALI's superiority over state-of-the-art competitors on simulated and real-world datasets.
Scientists at Southwest University create a Si3N4 microresonator to generate chip-scale microcombs with high nonlinearity, suitable for PRB generation. The application of chaotic optical frequency combs on random number generators could improve speed up to Pbits/s, offering low-cost and parallel solutions.
A team of researchers from East China University of Science and Technology has developed a systematic framework to identify indicator-PPCPs in raw landfill leachate samples. This study provides the first comprehensive analysis of 68 PPCPs in Shanghai, China landfill leachates, helping implement source apportionment in landfills.
Researchers propose using laser-induced photoreduction to decompose sphalerite and produce high-purity metal. The method eliminates greenhouse gas emissions and heavy metal pollution associated with traditional zinc electrolysis.
Researchers developed online machine learning models that can accurately predict wastewater influent flow rates under changing data patterns. The models outperformed conventional batch learning approaches, providing reliable tools for wastewater management and propelling the development of wastewater intelligence.
Researchers discovered that gut microbiota-derived 7-DHC regulates circadian rhythms, reduces inflammation in the gut, and treats experimental colitis by targeting key genes. This molecule has potential as a novel therapeutic agent for inflammatory bowel disease (IBD).
Researchers have developed a highly tunable mid-infrared laser based on the optical parametric oscillator (OPO) of BaGa4Se7, enabling high-resolution spectroscopy. The laser's wavelength tuning range spans 2.76-4.64 μm with a resolution of 0.3 nm.
A new method for treating wastewater generated from anaerobic digestion uses coal fly ash to remove pollutants, including COD and TP, with high efficiency. The treatment process also neutralizes pH and reduces heavy metal concentrations, making it suitable for irrigation water.
The study uses Aspen Plus software to establish a biomass gasification model for cow manure, finding optimal conditions for syngas production. The researchers evaluated parameters such as temperature, steam-to-biomass ratio, and pressure to enhance hydrogen content, while minimizing carbon dioxide and methane emissions.
Researchers found that combining alkaline pretreatment and air mixing improves methane yield kinetics and solids removal in co-digestion processes. This synergy promises even greater methane production and enhanced waste management efficiency.
Researchers simulated co-gasification of wood chip and potato peel to produce syngas, showing a positive synergistic interaction that increases carbon conversion efficiency. The study demonstrates the feasibility of this method for valorizing organic waste into valuable bioproducts while addressing environmental sustainability challenges.
Researchers have developed a novel membrane method to upgrade biogas performance using renewable aqueous ammonia solution, reducing CO2 and H2S impurities. The solution offers negligible CH4 loss and high reaction rate, making it an efficient alternative to traditional biogas upgrading technologies.
Researchers at HUST proposed using single sideband format in microwave communication systems to overcome photodetector bandwidth limitations. This approach significantly relaxes bandwidth restrictions, enabling higher communication speeds and microwave frequencies.
A real-time and accurate weight monitoring system was developed based on perching behavior, achieving an average accuracy of 99.5% and a variance in line with standard weights. The system demonstrated stability across different growth stages and rearing seasons.
Phototherapy during sleep effectively increases lymphatic excretion of beta-amyloid from brain tissues, cleansing the brain of toxic metabolites. This technique also demonstrates significant therapeutic effects over wakefulness, potentially offering new hope for Alzheimer's treatment.
Researchers developed a method to degrade organic pollutants and color removal in poultry litter digestate using photocatalytic titanate nanofibers. The study found optimal conditions for removing volatile fatty acids and chemical oxygen demand, demonstrating the effectiveness of TNFs in treating anaerobic digestate.
Researchers create BP/SnSe2-PVA composite to generate picosecond dissipative solitons, offering improved optical properties and tolerance to nonlinear effects. The composite material shows great potential for ultrafast optics applications.
The livestock and poultry industry in China is a significant source of greenhouse gas emissions, with over 3.8 Gt of manure produced annually. To achieve carbon neutrality, the country needs to implement efficient manure management systems that reduce emissions and promote renewable energy credits.