A5100715216- Natural Language Processing Techniques
- Topic Modeling
- Speech Recognition and Synthesis
- Anomaly Detection Techniques and Applications
- Chaos control and synchronization
- Advanced Graph Neural Networks
- Neural Networks Stability and Synchronization
- Speech and Audio Processing
- Network Security and Intrusion Detection
- Language, Metaphor, and Cognition
- Quantum chaos and dynamical systems
- Mathematical Dynamics and Fractals
- Nonlinear Dynamics and Pattern Formation
- Music and Audio Processing
- Neural Networks and Applications
- stochastic dynamics and bifurcation
- Privacy-Preserving Technologies in Data
- Syntax, Semantics, Linguistic Variation
- Environmental Impact and Sustainability
- Multimodal Machine Learning Applications
- Advanced Sensor and Control Systems
- Water Systems and Optimization
- Vibration and Dynamic Analysis
- Advanced Differential Equations and Dynamical Systems
- Graph Theory and Algorithms
China Southern Power Grid (China)
2024-2025
Naval University of Engineering
2023-2024
Huazhong University of Science and Technology
2008-2024
East China University of Science and Technology
2023
Yangtze University
2023
Zhejiang Chinese Medical University
2023
Dalian Polytechnic University
2023
Shandong Institute of Automation
2023
Chinese Academy of Sciences
2023
Xi'an Jiaotong University
2022
Exogenous microparticles including microplastics are novel pollutants that could persist in the environment with potential health effects, while crucial data on their exposure humans still lacking. To understand panorama of and distribution characteristics different kinds body fluids. A non-targeted microparticle internal landscape analysis was done thirteen human enclosed fluids covering eight systems. Totally 104 patients aged 24–96 years an average age 56 were included this study. After...
Speaker diarization is defined as the task of determining ldquowho spoke whenrdquo given an audio track and no other prior knowledge any kind. The following article shows how a state-of-the-art speaker system can be improved by combining traditional short-term features (MFCCs) with prosodic long-term features. First, we present framework to study discriminability 70 different Then, show top-ranked combined increase accuracy diarization. results were measured on standardized datasets (NIST...
This study intends to optimize the carbon footprint management model of power enterprises through artificial intelligence (AI) technology help scientific formulation emission reduction strategies. Firstly, a calculation based on big data and AI is established, then machine learning algorithm used deeply mine identify main influencing factors opportunities. Finally, driver-state-response (DSR) evaluate audit industry comprehensively analyze effect reduction. Taking China Electric Power...
During the past few years, speaker diarization has achieved satisfying accuracy in terms of Diarization Error Rate (DER). The most successful approaches, based on agglomerative clustering, however, exhibit an inherent computational complexity which makes real-time processing, especially combination with further processing steps, almost impossible. In this article we present a framework to speed up clustering diarization. basic idea is adopt computationally cheap method reduce hypothesis...
Graph-structured data arise in wide applications, such as computer vision, bioinformatics, and social networks. Quantifying similarities among graphs is a fundamental problem. In this paper, we develop framework for computing graph kernels, based on return probabilities of random walks. The advantages our proposed kernels are that they can effectively exploit various node attributes, while being scalable to large datasets. We conduct extensive classification experiments evaluate kernels....
We propose a novel neural label embedding (NLE) scheme for the domain adaptation of deep network (DNN) acoustic model with unpaired data samples from source and target domains. With NLE method, we distill knowledge powerful source-domain DNN into dictionary embeddings, or l-vectors, one each senone class. Each l-vector is representation senone-specific output distributions learned to minimize average L <sub xmlns:mml="http://www.w3.org/1998/Math/MathML"...
The goal of vocabulary optimization is to construct a with exactly those words that are the most likely appear in test data. We will present new approach reduce out-of-vocabulary (OOV) rate by adapting model during ASR process. This method can also be used for statistical language (SLM) adaptation. An information retrieval system after first pass obtain set relevant documents. These documents then generate and/or corpus. In this paper, we propose retrieving well-adapted purpose....
Chronic pain is a common long-term condition that affects person's physical and emotional functioning. Currently, the integrated biopsychosocial approach mainstay treatment for people with chronic pain. Self-reporting (the use of questionnaires) one most methods to evaluate outcome. The questionnaires can consist more than 300 questions, which tedious complete at home. This paper presents machine learning analyze self-reporting data collected from treatment, in order identify an optimal set...
In Western countries, women and men, young old, enroll in adult education for different reasons. This is even more the case China. study helps Shanghai program planners better appreciate learners by understandinghow motivational orientations vary as a function of gender age. The Chinese version Education Participation Scale was administered to 448 adults enrolled classes. It had factor structure loosely comparable English version. Women were inclined than men be social stimulation, contact...
In this letter we report a new class of chaotic three-neuron cellular neural networks that are described by special three-dimensional autonomous ordinary differential equations.
Traditional signature-based detection methods fail to detect unknown malwares, while data mining for are proved useful new malwares but suffer high false positive rate. In this paper, we provide a novel hybrid framework called HRS based on the analysis 50 millions of malware samples across 20,000 classes from our antivirus platform. The distribution elaborated and is proposed, which consists Hash-based, Rule-based SVM-based models trained different according distribution. model core...
Language policy, which is directly concerned with language practice, ideology and management, has become increasingly important in real social life. Explicit policies different fields, such as texts law, education, the public, have been explored for many years. However, global comparative research on various constitutional (CT) quite limited. In response, present study aimed to investigate framework features of CT through a combination qualitative quantitative analysis. Results showed that:...
In hydraulic power systems with water valves, cavitation is more likely to happen than in an oil system. A two-step throttle, which consists of two orifices, effective way avoid or reduce damage. this research, experiments are conducted investigate the flow-pressure and characteristics throttles different shapes. The working medium tap water. new indicator introduced detect inception, namely that a particular low-frequency component outlet pressure fluctuation valve appears grows rapidly...
With the advance of language models, privacy protection is receiving more attention. Training data extraction therefore great importance, as it can serve a potential tool to assess leakage. However, due difficulty this task, most existing methods are proof-of-concept and still not effective enough. In paper, we investigate benchmark tricks for improving training using publicly available dataset. Because use pipeline generating-then-ranking, i.e., generating text candidates then ranking them...