A5116294179- Topic Modeling
- Domain Adaptation and Few-Shot Learning
- Advanced Image and Video Retrieval Techniques
- Multimodal Machine Learning Applications
- Data Management and Algorithms
- Luminescence Properties of Advanced Materials
- Image Processing Techniques and Applications
- Microfluidic and Bio-sensing Technologies
- Human Pose and Action Recognition
- Advanced Algorithms and Applications
- Advanced Graph Neural Networks
- CAR-T cell therapy research
- UAV Applications and Optimization
- Artificial Intelligence in Games
- Flood Risk Assessment and Management
- Hydrological Forecasting Using AI
- Reinforcement Learning in Robotics
- Caching and Content Delivery
- Engineering and Test Systems
- Hydrology and Sediment Transport Processes
- Data Stream Mining Techniques
- Advancements in Photolithography Techniques
- Machine Learning and Algorithms
- Micro and Nano Robotics
- Control Systems in Engineering
Changjiang Water Resources Commission
2024-2025
Sichuan University
2023-2024
Chengdu University
2023-2024
Zhejiang University
2023
Anhui University of Technology
2023
Innovation Team (China)
2023
Microsoft (United States)
2020-2021
Qingdao University
2021
Microsoft Research Asia (China)
2020-2021
Soochow University
2018-2019
Obtaining single cell from tissues is important for intersection of information and bioscience research. In this letter, a robotic framework based on micromanipulation system has been proposed, which can automatically intelligently cut down cells intact tissue sections. The proposed method consists several steps. An attention mechanism improved (AMI) tip localization neural network to detect track the needle micro-scale displacement end-effector within limited field view under microscopy....
Decitabine, a promising epi-immunotherapeutic agent has shown clinical responses in solid tumor patients, while the anti-tumor mechanisms were unclear. We aimed to investigate immunomodulatory effect of decitabine peripheral T cells.We applied next-generation sequencing complementarity-determining region 3 (CDR3) TCRβ gene, diversity which acts as prerequisite for host immune system recognize universal foreign antigens. collected blood mononuclear cells (PBMCs) from 4 at baseline and after 2...
Anti-CD19 chimeric antigen receptor-modified T (CAR-T-19) cells have emerged as a powerful targeted immunotherapy for B-cell lineage acute lymphoblastic leukemia with remarkable clinical response in recent trials. Nonetheless, few data are available on the subsequent monitoring and treatment of patients, especially those disease recurrence after CAR-T-19 cell infusion. Here, we analyzed three patients who survived our phase I trial were studied by means biomarkers reflecting persistence vivo...
In recent years, as river-borne sediment transport has decreased, land loss in estuarine areas become increasingly severe. Particularly from 2009 to 2022, due the reduced load upper reaches of Yangtze River, significant changes erosion-deposition conditions at River Estuary have a critical concern. The suspended concentration (SSC) estuary is closely linked content, and against this backdrop, study aims thoroughly explore response mechanisms Estuary's SSC conditions, with goal providing...
Effectively simulating the variation in suspended sediment concentration (SSC) estuaries during typhoons is significant for water quality and ecological conditions of estuarine shoal wetlands their adjacent coastal waters. During typhoons, SSC undergoes large variations due to changes meteorological hydrological factors such as waves, wind speed, precipitation, which increases difficulty SSC. Therefore, this study, we use an optimized Principal Component Analysis Long Short-Term Memory...
The impairment of angiogenesis is an outstanding pathogenic characteristic glucocorticoid (GC)-induced osteonecrosis the femoral head (ONFH). Human umbilical cord mesenchymal stem cells (hUC-MSCs) have been used in several diseases models, which were reported to be involved angiogenesis. However, whether hUC-MSCs suppress GC-induced ONFH via promoting still unclear. isolated from Wharton's jelly using explant culture method. A model was established vitro and vivo. angiogenesis, proliferation...
The in-memory algorithms for approximate nearest neighbor search (ANNS) have achieved great success fast high-recall search, but are extremely expensive when handling very large scale database. Thus, there is an increasing request the hybrid ANNS solutions with small memory and inexpensive solid-state drive (SSD). In this paper, we present a simple efficient memory-disk indexing system, named SPANN, that follows inverted index methodology. It stores centroid points of posting lists in disk....
Deep matching models aim to facilitate search engines retrieving more relevant documents by mapping queries and into semantic vectors in the first-stage retrieval. When leveraging BERT as deep model, attention score across two words are solely built upon local contextualized word embeddings. It lacks prior global knowledge distinguish importance of different words, which has been proved play a critical role information retrieval tasks. In addition this, only performs sub-words tokens weakens...
Cell trapping is considered a complicated and challenging task due to the small scale of manipulation involved. Image guided robotic micro among novel promising solutions that may enhance human capabilities. In this paper we demonstrate 3-degree-of-freedom (3-DOF) wireless magnetic control an untethered microrobot (2-DOF position 1-DOF pointing orientation). An accurate orientation model has been employed determine microrobot's in 2D space. addition, multiple experiments were conducted...
Quantum circuit mapping (QCM) is a crucial preprocessing step for executing logical (LC) on noisy intermediate-scale quantum (NISQ) devices. Balancing the introduction of extra gates and efficiency poses significant challenge process. To address this challenge, we propose mutual-influence-aware (MIA) heuristic method by integrating an initial search framework, generator, mapper. Initially, framework utilizes generator to obtain favorable starting point search. With point, process can...
Many LLM tasks are performed in large batches or even offline, and the performance indictor for which is throughput. These usually show characteristic of prefix sharing, where different prompt input can partially common prefix. However, existing inference engines tend to optimize streaming requests limitations supporting batched with sharing characteristic. The solutions use LRU-based cache reuse KV context that about be reused may prematurely evicted implicit management. Even if not...
Quantization has become one of the most effective methodologies to compress LLMs into smaller size. However, existing quantization solutions still show limitations either non-negligible accuracy drop or system inefficiency. In this paper, we make a comprehensive analysis general principles on their effect triangle accuracy, memory consumption and efficiency. We propose MixLLM that explores new optimization space mixed-precision between output features based insight different matter...
To achieve good result quality and short query response time, search engines use specific match plans on Inverted Index to help retrieve a small set of relevant documents from billions web pages. A plan is composed sequence rules, which contain discrete rule types continuous stopping quotas. Currently, are manually designed by experts according their several years' experience, encounters difficulty in dealing with heterogeneous queries varying data distribution. In this work, we formulate...