A5100297472- Air Quality Monitoring and Forecasting
- Advanced biosensing and bioanalysis techniques
- Advanced Memory and Neural Computing
- Atmospheric chemistry and aerosols
- Ferroelectric and Negative Capacitance Devices
- Water Quality and Pollution Assessment
- Biosensors and Analytical Detection
- Bacteriophages and microbial interactions
- Machine Learning and ELM
- Atmospheric and Environmental Gas Dynamics
- Innovative Microfluidic and Catalytic Techniques Innovation
- dental development and anomalies
- Biocrusts and Microbial Ecology
- Regional Development and Environment
- Advanced Biosensing Techniques and Applications
- Orthodontics and Dentofacial Orthopedics
- Machine Learning and Data Classification
- Healthcare and Venom Research
- Cloud Computing and Resource Management
- Proteoglycans and glycosaminoglycans research
- Long-Term Effects of COVID-19
- Gaussian Processes and Bayesian Inference
- Viral gastroenteritis research and epidemiology
- Big Data Technologies and Applications
- Advanced Nanomaterials in Catalysis
Berkeley College
2025
University of California, Berkeley
2025
Sun Yat-sen University
2024
Wuhan University
2024
Shihezi University
2023
Suzhou Institute of Biomedical Engineering and Technology
2021-2023
Chinese Academy of Sciences
2022-2023
Shanghai Jiao Tong University
2021-2023
State Key Joint Laboratory of Environment Simulation and Pollution Control
2023
Tsinghua University
2023
Diffusion Transformers (DiTs) dominate video generation but their high computational cost severely limits real-world applicability, usually requiring tens of minutes to generate a few seconds even on high-performance GPUs. This inefficiency primarily arises from the quadratic complexity 3D Full Attention with respect context length. In this paper, we propose training-free framework termed Sparse VideoGen (SVG) that leverages inherent sparsity in boost inference efficiency. We reveal...
Vegetation pattern generated by aeolian sand movements is a typical type of vegetation in arid and semi-arid areas. Considering the effects on diffusion as cross-diffusion term, an improved vegetation-sand studied to explore formation transition. Turing bifurcation analysis amplitude are performed. Theoretical shows that three types patterns can be formed: stripes, hexagons mixture stripes hexagons. Numerical simulations consistent with theoretical results, which also show hysteresis...
Resistive Random-Access-Memory (ReRAM) cross-bar is a promising technique for deep neural network (DNN) accelerators, thanks to its in-memory and in-situ analog computing abilities Vector-Matrix Multiplication-and-Accumulations (VMMs). However, it challenging crossbar architecture exploit the sparsity in DNNs. It inevitably causes complex costly control fine-grained due limitation of tightly-coupled structure.As countermeasure, we develop novel ReRAM-based DNN accelerator, named...
Abstract Identification, sorting, and sequencing of individual cells directly from in situ samples have great potential for in‐depth analysis the structure function microbiomes. In this work, based on an artificial intelligence (AI)‐assisted object detection model cell phenotype screening a cross‐interface contact method single‐cell exporting, we developed automatic index‐based system called EasySort AUTO, where microbial are sorted then packaged microdroplet automatically exported precisely...
Antibiotic resistance is a pressing global health issue, leading to increased illnesses and fatalities. The contribution of viruses the acquisition, preservation, dissemination antibiotic genes (ARGs) not yet fully understood. By using high-throughput functional gene-based microarray (GeoChip 5.0), this study examines prevalence relative abundance bacteriophage eukaryotic viral in swine manure, compost, compost-amended agricultural soil, unamended soil from suburban regions Beijing, China....
Abstract. The acid deposition has been considered to be a severe environmental issue in China. pH, electrical conductivity (EC), and the concentrations of water soluble ions (NO3−, Cl−, Ca2+, K+, F−, NH4+, Mg2+, SO42−, Na+) precipitation samples collected from 320 cities during 2011–2016 across whole China were measured. mean NO3− SO42− order winter (6.10, 19.44 45.74 μeq/L) > spring (3.45, 13.83, 42.61 autumn (2.67, 9.73, 28.85 summer (2.04, 7.66, 19.26 μeq/L). secondary (SO42−, NH4+),...
Glioblastoma multiforme (GBM) is a prevalent intracranial brain tumor associated with high rate of recurrence and treatment difficulty. The prediction novel molecular biomarkers through bioinformatics analysis may provide new clues into early detection eventual GBM. Here, we used data from the GTEx TCGA databases to identify 1923 differentially expressed genes (DEGs). GO KEGG analyses indicated that DEGs were significantly enriched in immune response coronavirus disease-COVID-19 pathways....
pH was one of the important meat quality traits, which an factor affecting storage/shelf life and in production. In order to find a way extend life, values (pH45min, pH24h, pH48h pH72h) longissimus dorsi muscles F2 individuals 462 Texel sheep × Altay were determined, genotyping performed using Illumina Ovine SNP 600 K BeadChip whole genome resequencing technology, genome-wide association analysis (GWAS) used screen candidate genes molecular markers for related traits mutton, effects...
The photogenerated charge carrier separation and transportation of inside photocathodes can greatly influence the performance photoelectrochemical (PEC) H
Combined orthognathic-orthodontic treatment can be lengthy and expensive. It is therefore important to evaluate the effectiveness of proposed likely outcomes.To determine demographic surgical details patients who received at Christchurch Hospital, New Zealand, assess using Severity Outcome Index (SOI).An observational retrospective study was conducted orthodontic between 2005 2012 Hospital. Pre- post-treatment lateral cephalometric radiographs 93 were evaluated. Seven parameters assessed...
As the demand for long-context large language models (LLMs) increases, with context windows of up to 128K or 1M tokens are becoming increasingly prevalent. However, LLM inference is challenging since speed decreases significantly as sequence length grows. This slowdown primarily caused by loading a KV cache during self-attention. Previous works have shown that small portion critical will dominate attention outcomes. we observe criticality token highly depends on query. To this end, propose...
Periodontitis, with its persistent nature, causes significant distress for most sufferers. Current treatments, such as mechanical cleaning and surgery, often fail to fully address the underlying overactivation of fibroblasts that drives this degradation. Targeting post-transcriptional regulation fibroblasts, particularly at 3′-untranslated regions (3′UTR) pathogenic genes, offers a therapeutic strategy periodontitis. Herein, we developed DNA nanorobot purpose. This system uses dynamic...
Offline batch inference, which leverages the flexibility of request batching to achieve higher throughput and lower costs, is becoming more popular for latency-insensitive applications. Meanwhile, recent progress in model capability modality makes requests diverse compute memory demands, creating unique opportunities improvement by resource overlapping. However, a schedule that maximizes overlapping can conflict with prefix sharing, widely-used performance optimization, causing sub-optimal...
There is a close relationship between architectural color and the formation of urban social historical context, features, residents' attachment, so an important part features.The paper generalizes features during various stages development in Changchun, analyzes main factors influencing Changchun from four aspects which are natural environment, traditional regional culture, modern culture building technology material, proposes planning proposals provides reference for extending cultural feature.
The second-order training methods can converge much faster than first-order optimizers in DNN training. This is because the utilizes inversion of information (SOI) matrix to find a more accurate descent direction and step size. However, huge SOI matrices bring significant computational memory overheads traditional architectures like GPU CPU. On other side, ReRAM-based process-in-memory (PIM) technology suitable for following three reasons: First, PIM's computation happens memory, which...
Resistive Random-Access-Memory (ReRAM) crossbar is a promising technique for deep neural network (DNN) accelerators, thanks to its in-memory and in-situ analog computing abilities Vector-Matrix Multiplication-and-Accumulations (VMMs). However, it challenging architecture exploit the sparsity in DNN. It inevitably causes complex costly control fine-grained due limitation of tightly-coupled structure. As countermeasure, we developed novel ReRAM-based DNN accelerator, named...
The detection of multiplex cytokines has attracted considerable attention due to its enormous value in the fields biology and medicine. In this research, we proposed a host-guest structure bifunctional dextran microsphere synthesized by an emulsion polymerization method microfluidic device. was composed magnetic microspheres (MDM) as host multiple fluorescent molecules guest. Fe3O4 nanoparticles were encapsulated into microsphere. Cyanine5 NHS ester azide (Cy5) Cyanine2 (Cy2) grafted onto...