A5079750750- Topic Modeling
- Speech and dialogue systems
- Natural Language Processing Techniques
- Speech Recognition and Synthesis
- Speech and Audio Processing
- Traffic Prediction and Management Techniques
- Human Mobility and Location-Based Analysis
- Advanced Graph Neural Networks
- Multimodal Machine Learning Applications
- Music and Audio Processing
- Domain Adaptation and Few-Shot Learning
- AI in Service Interactions
- Text and Document Classification Technologies
- Recommender Systems and Techniques
- Sentiment Analysis and Opinion Mining
- Complex Network Analysis Techniques
- Advanced Text Analysis Techniques
- Transportation Planning and Optimization
- Service-Oriented Architecture and Web Services
- Web Data Mining and Analysis
- Multi-Agent Systems and Negotiation
- Data Quality and Management
- Semantic Web and Ontologies
- Neural Networks and Applications
- Vehicular Ad Hoc Networks (VANETs)
China Mobile (China)
2016-2025
Tsinghua University
2021-2024
Tongji University
2018-2021
AT&T (United States)
2004-2021
Shanghai Tenth People's Hospital
2018-2021
Fudan University
2018
Institute of Acoustics
2002-2016
Chinese Academy of Sciences
2016
Shanghai Sixth People's Hospital
2014-2016
Shanghai Jiao Tong University
2014-2016
// Yongzhi Yang 1,* , Xia 2,* Hongqi Chen 2 Leiming Hong 1 Junlan Feng Jun Zhe Chenzhang Shi Wen Wu Renyuan Gao Qing Wei 3 Huanlong Qin and Yanlei Ma Department of GI Surgery, Shanghai Tenth People's Hospital Affiliated to Tongji University, Shanghai, China Jiao Tong University Sixth Hospital, Pathology, * These authors have contributed equally this work Correspondence to: Qin, email: Ma, Wei, Keywords : colorectal cancer, perioperative probiotics treatment, post-operative outcome,...
Social influence can be described as power - the ability of a person to thoughts or actions others. Identifying influential users on online social networks such Twitter has been actively studied recently. In this paper, we investigate modified k-shell decomposition algorithm for computing user Twitter. The input is connection graph between defined by follower relationship. User measured level, which output algorithm. Our first insight modify assign logarithmic values users, producing measure...
Fine-grained image recognition is a longstanding computer vision challenge that focuses on differentiating objects belonging to multiple subordinate categories within the same meta-category. Since images meta-category usually share similar visual appearances, mining discriminative cues key distinguishing fine-grained categories. Although commonly used <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">image-level</i> data augmentation techniques...
Our previously published study documented a deregulation of the microRNA miR-150 in colorectal cancer. Here, we investigated further, vitro and vivo, potential molecular mechanisms underlying involvement cancer, using appropriate biological methods. We report that is key regulator tumourigenesis progression by acting as tumour suppressor targeting c-Myb. The current findings suggest may have important roles pathogenesis
Accurately predicting base station traffic volumes and understanding mobile patterns is essential for smart city development, enabling efficient resource allocation ensuring high-quality communication services. However, existing works have limitations in capturing spatial information, though the surrounding environment plays a critical role prediction. In this paper, we utilize knowledge graph to represent information add important urban components augment it making more effective tool...
We present a machine learning approach to sentiment classification on twitter messages (tweets). classify each tweet into two categories: polar and non-polar. Tweets with positive or negative are considered polar. They non-polar otherwise. Sentiment analysis of tweets can potentially benefit different parties, such as consumers marketing researchers, for obtaining opinions products services. methods text normalization the noisy their respect polarity. experiment mixture model generation...
Structured belief states are crucial for user goal tracking and database query in task-oriented dialog systems. However, training trackers often requires expensive turn-level annotations of every utterance. In this paper we aim at alleviating the reliance on state labels building end-to-end systems, by leveraging unlabeled data towards semi-supervised learning. We propose a probabilistic model, called LAtent BElief State (LABES) where represented as discrete latent variables jointly modeled...
piRNA-823 as a member of the piRNA family is reported to promote tumour cell proliferation in multiple myeloma and hepatocellular cancer. However, few studies on function colorectal cancer (CRC). Our present study data showed that plays an oncogene role CRC cells. Inhibition can significantly inhibit proliferation, invasion apoptosis resistance Mechanism have shown inhibits ubiquitination hypoxia-inducible factor-1 alpha (HIF-1α) by up-regulating expression Glucose-6-phosphate dehydrogenase...
Topology impacts important network performance metrics, including link utilization, throughput and latency, is of central importance to operators. However, due the combinatorial nature topology, it extremely difficult obtain an optimal solution, especially since topology planning in networks also often comes with management-specific constraints. As a result, local optimization hand-tuned heuristic methods from human experts adopted practice. Yet, cannot cover global design space while taking...
Early exiting has become a promising approach to improving the inference efficiency of deep networks. By structuring models with multiple classifiers (exits), predictions for "easy" samples can be generated at earlier exits, negating need executing deeper layers. Current multi-exit networks typically implement linear intermediate layers, compelling low-level features encapsulate high-level semantics. This sub-optimal design invariably undermines performance later exits. In this paper, we...
A Dialogue State Tracker (DST) is a core component of modular task-oriented dialogue system. Tremendous progress has been made in recent years. However, the major challenges remain. The state-of-the-art accuracy for DST below 50% multi-domain task. learnable any new domain requires large amount labeled in-domain data and training from scratch. In this paper, we propose Meta-Reinforced Multi-Domain Generator (MERET). Our first contribution to improve accuracy. We enhance neural model based...
Self-supervised entity alignment (EA) aims to link equivalent entities across different knowledge graphs (KGs) without the use of pre-aligned pairs. The current state-of-the-art (SOTA) self-supervised EA approach draws inspiration from contrastive learning, originally designed in computer vision based on instance discrimination and loss, suffers two shortcomings. Firstly, it puts unidirectional emphasis pushing sampled negative far away rather than pulling positively aligned pairs close, as...
Graphs are widely used to represent the relations among entities. When one owns complete data, an entire graph can be easily built, therefore performing analysis on is straightforward. However, in many scenarios, it impractical centralize data due privacy concerns. An organization or party only keeps a part of whole i.e., isolated from different parties. Recently, Federated Learning (FL) has been proposed solve isolation issue, mainly for Euclidean data. It still challenge apply FL because...
Recently, two approaches, fine-tuning large pre-trained language models and variational training, have attracted significant interests, separately, for semi-supervised end-to-end task-oriented dialog (TOD) systems. In this paper, we propose Variational Latent-State GPT model (VLS-GPT), which is the first to combine strengths of approaches. Among many options models, generative inference learning TOD system, both as auto-regressive based on GPT-2, can be further trained over a mix labeled...
With the rapid development of cellular network, network planning is increasingly important. Generating large-scale urban traffic contributes to via simulating behaviors planned network. Existing methods fail in long-term temporal while cannot model influences environment on networks. We propose a knowledge-enhanced GAN with multi-periodic patterns generate based environment. First, we design simulate and aperiodic dynamics learning daily patterns, weekly residual between periodic step by...
Knowledge graphs are widely used in industrial applications, making error detection crucial for ensuring the reliability of downstream applications. Existing methods often fail to effectively leverage fine-grained subgraph information and rely solely on fixed graph structures, while also lacking transparency their decision-making processes, which results suboptimal performance. In this paper, we propose a novel Multi-Agent framework Graph Error Detection (MAKGED) that utilizes multiple large...
Numerical investigations aiming to determine the universality class of critical motility-induced phase separation (MIPS) in two dimensions (2D) have resulted inconclusive findings. Here, using finite-size scaling results obtained from large-scale computer simulations, we find that static and dynamic exponents associated with 3D MIPS all closely match those Ising a conserved scalar order parameter. This finding is corroborated by fluctuating hydrodynamic description dynamics parameter field...
With the rapid advancement of Large Language Models (LLMs), safety LLMs has been a critical concern requiring precise assessment. Current benchmarks primarily concentrate on single-turn dialogues or single jailbreak attack method to assess safety. Additionally, these have not taken into account LLM's capability identifying and handling unsafe information in detail. To address issues, we propose fine-grained benchmark SafeDialBench for evaluating across various attacks multi-turn dialogues....