A5101671391- Domain Adaptation and Few-Shot Learning
- Generative Adversarial Networks and Image Synthesis
- Multimodal Machine Learning Applications
- Natural Language Processing Techniques
- Human Pose and Action Recognition
- Advanced Memory and Neural Computing
- Topic Modeling
- CCD and CMOS Imaging Sensors
- Evacuation and Crowd Dynamics
- AI in cancer detection
- Liver Disease Diagnosis and Treatment
- Anomaly Detection Techniques and Applications
- Hops Chemistry and Applications
- Mobile Health and mHealth Applications
- FinTech, Crowdfunding, Digital Finance
- Diabetes and associated disorders
- PI3K/AKT/mTOR signaling in cancer
- Subtitles and Audiovisual Media
- Energy Load and Power Forecasting
- HER2/EGFR in Cancer Research
- Evaluation Methods in Various Fields
- Flow Measurement and Analysis
- Hand Gesture Recognition Systems
- Infrared Target Detection Methodologies
- Advanced Algorithms and Applications
Sichuan University
2024
Heilongjiang University
2024
Nanjing University of Finance and Economics
2023-2024
King University
2024
Nanchang University
2024
Peking University
2024
Nanyang Technological University
2024
Nanjing University
2023
Massey University
2023
North University of China
2013
Diffusion models have achieved great success in image synthesis through iterative noise estimation using deep neural networks. However, the slow inference, high memory consumption, and computation intensity of model hinder efficient adoption diffusion models. Although post-training quantization (PTQ) is considered a go-to compression method for other tasks, it does not work out-of-the-box on We propose novel PTQ specifically tailored towards unique multi-timestep pipeline architecture...
The complicated architecture and high training cost of vision transformers urge the exploration post-training quantization. However, heavy-tailed distribution transformer activations hinders effectiveness previous quantization methods, even with advanced quantizer designs. Instead tuning to better fit activation distribution, this paper proposes NoisyQuant, a quantizer-agnostic enhancement for performance transformers. We make surprising theoretical discovery that given quantizer, adding...
Review Application of Wearable Devices in Diabetes Management Zijing Du 1,2,†, Feifan Zhang 1,†, Yifei Ge 1, Yijiang Liu 3, Honghua Yu 2, Yong Wang 4, Rinkoo Dalan 1,5, and Xiaotao Shen 1,3,* 1 Lee Kong Chian School Medicine, Nanyang Technological University, Singapore, 636921, Singapore 2 Guangdong Eye Institute, Department Ophthalmology, Provincial People’s Hospital (Guangdong Academy Medical Sciences), Southern Guangzhou, 510080, China 3 Chemistry, Chemical Engineering Biotechnology,...
Introduction Inflammation is a key driver of adverse outcomes in acute myocardial infarction (AMI), yet current western anti-inflammatory therapies are limited by their single-target nature and side effects. Traditional Chinese medicine (TCM), such as Tanhuo Decoction (THD), offers multi-target, low-toxicity alternative. Methods In randomized controlled trial, AMI patients with high inflammatory responses received either standard Western (WM) alone or combined THD for 3 days. Clinical...
Severe acute pancreatitis (SAP), a widespread inflammatory condition impacting the abdomen with high mortality rate, poses challenges due to its unclear pathogenesis and absence of effective treatment options. Isorhamnetin (ISO), naturally occurring flavonoid, demonstrates robust antioxidant anti-inflammatory properties intricately linked modulation mitochondrial function. However, specific protective impact ISO on SAP remains be fully elucidated. In this study, we demonstrated that...
In the task of monocular 3D pose estimation, estimation errors limb joints (i.e., wrist, ankle, etc) with a higher degree freedom(DOF) are larger than that others hip, thorax, etc). Specifically, may accumulate along physiological structure human body parts, and trajectories DOF bring in complexity. To address this problem, we propose aware framework, involving kinematic constraint network as well trajectory temporal module, to improve prediction accuracy joint positions. Two constraints...
In order to eliminate the sensor's dynamic error, a method that improved recursive least squares is proposed. Gradient descent used generate initial value of filter parameters, then (RLS) algorithm optimize parameters. The tested and verified on matlab platform. time-domain response frequency domain sensor are analyzed before after compensation. piezoelectric CY_YD-205 compensated. Engineering experiments show compensation designed by can improve characteristics system.
Large Language Models (LLMs) have demonstrated significant potential in performing multiple tasks multimedia applications, ranging from content generation to interactive entertainment, and artistic creation. However, the diversity of downstream multitask scenarios presents substantial adaptation challenges for LLMs. While traditional methods often succumb knowledge confusion on their monolithic dense models, Mixture-of-Experts (MoE) has been emerged as a promising solution with its sparse...
Abstract Hyperlipidemia presents a significant global healthcare challenge, necessitating innovative therapeutic strategies for more effective outcomes. Recent studies have highlighted the beneficial impact of moderate beer intake on metabolic diseases. The purpose this research is to explore possible molecular mechanisms isoxanthohumol (IXN), major hop flavonoid in beer, treatment hyperlipidemia. mice model acute hyperlipidemia was constructed by intraperitoneal injection Triton WR‐1339....
This study aims to investigate, through neuroscientific methods, the effects of particular architectural elements on pedestrian spatial cognition and experience in analysis design walking street spaces. More precisely, this paper will describe impact density variation storefront signs brainwaves passersby East Asian city streets, providing strategies guidelines for urban development renewal. Firstly, summarizes research method review questions related literature; secondly, establishes...
Esophageal squamous cell carcinoma (ESCC) is a particularly aggressive form of cancer with high mortality. In the present study, novel 8‑hydroxyquinoline derivative (91b1) was investigated for its anticancer activities in ESCC along associated mechanisms. The
Disulfidptosis, which was recently identified, has shown promise as a potential cancer treatment. Nonetheless, the precise role of long non‑coding RNAs (lncRNAs) in this phenomenon is currently unclear. To elucidate their significance bladder (BLCA), signature disulfidptosis‑related lncRNAs (DRlncRNAs) developed and prognostic explored. BLCA sample data were sourced from The Cancer Genome Atlas. A predictive comprising DRlncRNAs formulated subsequently validated. combination with clinical...
Abstract As China’s economy enters a stage of high-quality development, the trend power demand has changed and can affect investment construction industry. A series key criteria impacting on were sorter out to build index system currently. The time difference correlation analysis was applied classify these as precursor, consistency lag index. based three categories fitted multivariate statistical analysis. Then comprehensive assembled forecast electric from 2020 2022 in China current policy...
Diffusion models have achieved great success in image synthesis through iterative noise estimation using deep neural networks. However, the slow inference, high memory consumption, and computation intensity of model hinder efficient adoption diffusion models. Although post-training quantization (PTQ) is considered a go-to compression method for other tasks, it does not work out-of-the-box on We propose novel PTQ specifically tailored towards unique multi-timestep pipeline architecture...
As the capabilities of Large-Language Models (LLMs) become widely recognized, there is an increasing demand for human-machine chat applications. Human interaction with text often inherently invokes mental imagery, aspect that existing LLM-based chatbots like GPT-4 do not currently emulate, as they are confined to generating text-only content. To bridge this gap, we introduce ChatIllusion, advanced Generative multimodal large language model (MLLM) combines LLM only visual comprehension but...
The burgeoning field of Multimodal Large Language Models (MLLMs) has exhibited remarkable performance in diverse tasks such as captioning, commonsense reasoning, and visual scene understanding. However, the deployment these large-scale MLLMs on client devices is hindered by their extensive model parameters, leading to a notable decline generalization capabilities when models are compressed for device deployment. Addressing this challenge, we introduce Cloud-Device Collaborative Continual...
With the rapid development of network, World Wide Web has become carrier a large amount information, which contains lot valuable and its great potential can not be underestimated. At present, both at home abroad are paying attention to business opportunities brought by big data, all walks life have their own data achieve greatest commercial profits in market. processing technology storage system advantages, but it also shortcomings large-scale processing. This article takes transaction as...
The complicated architecture and high training cost of vision transformers urge the exploration post-training quantization. However, heavy-tailed distribution transformer activations hinders effectiveness previous quantization methods, even with advanced quantizer designs. Instead tuning to better fit activation distribution, this paper proposes NoisyQuant, a quantizer-agnostic enhancement for performance transformers. We make surprising theoretical discovery that given quantizer, adding...