A5101614636- Neural Networks and Applications
- Higher Education and Teaching Methods
- Speech Recognition and Synthesis
- Educational Technology and Pedagogy
- Educational Technology and Assessment
- Speech and Audio Processing
- Advanced Decision-Making Techniques
- EFL/ESL Teaching and Learning
- Model Reduction and Neural Networks
- Gene expression and cancer classification
- Cell Image Analysis Techniques
- AI in cancer detection
- Rough Sets and Fuzzy Logic
- Fault Detection and Control Systems
- Second Language Acquisition and Learning
- Advanced Computational Techniques and Applications
- Neural Networks and Reservoir Computing
- Evaluation and Optimization Models
- Scientific Research and Technology
- Online Learning and Analytics
- Infrared Target Detection Methodologies
- Indoor and Outdoor Localization Technologies
- Advanced Algorithms and Applications
- Face recognition and analysis
- Perfectionism, Procrastination, Anxiety Studies
Southwest Jiaotong University
2025
Xihua University
2024
Wuhan University
2024
Tianjin University
2014
Huanggang Normal University
2012
Mianyang Normal University
2009
Shanghai Library
2004
Saudi Heart Association
2003
University of Maryland, College Park
2002-2003
Lanzhou Petrochemical Polytechnic
2003
This study explores how graduate students' mentorship homegate (or team) support (GSMTS) and challenging-hindering pressures impact their intrinsic motivation for research, identification with research roles, innovative behaviors. Data from 548 students were collected using convenience sampling analyzed Amos SPSS statistical software package via questionnaires distributed to universities in SiChuan province of China. The findings reveal that (1) stress can not directly positively predict...
Two-dimensional (2-D) direction-of-arrival (DOA) estimation method using three-parallel uniform linear arrays (ULAs) is proposed in this letter. The 2-D DOA problem addressed by making full use of elements the ULAs. Furthermore, algorithm has better angle performance practical mobile elevation angles between 70° and 90° can automatically pair estimated azimuth with lower complexity. Simulation results demonstrate effectiveness algorithm.
In principle, a potentially infinitely large neural network (either in number of neurons or the precision single activity) could possess an equivalent computational power to Turing machine. The authors show such equivalence machines several explicitly constructed networks. It is proven that for any given machine there exists recurrent with local, second-order, and uniformly connected weights (i.e., connecting second-order product local 'input neurons' their corresponding 'output neurons')...
A dynamic time warping based speech recognition system with neural network trained templates is proposed. The algorithm for training the derived on minimizing classification error of classifier. speaker-independent isolated digit experiment conducted and achieves a 0.89% average rate only one template each digit, indicating that are able to capture speaker-invariant features signals. Both nondiscriminative discriminative versions net considered. former maximum likelihood estimation. latter...
A discriminative training algorithm for predictive neural network models is proposed. The applied to a speaker independent isolated digit recognition experiment. error rate reduced from 2.52% when the classifier trained with non-discriminative 0.58% applied. increase in discrimination ability also demonstrated.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
The authors present arguments for a fast online training algorithm recurrent neural networks. They an that would require O(N/sup 3/) calculations to update the weights in one time step, which is faster than all other known algorithms. formulate derivations of this variational approach has advantage providing unified view various algorithms were derived by number researchers from very different circumstances.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML"...
An adaptive template method for pattern recognition is proposed. The adaptation algorithm derived based on minimizing the classification error of classifier. authors have applied this to a multispeaker English E-set experiment and achieved 90.38% average rate with only one each letter. This indicates that templates are able capture speaker-invariant features speech signals.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Ponte Academic JournalNov 2017, Volume 73, Issue 11 TEACHING SCHEME INNOVATION OF ACCOUNTING COURSES UNDER INTERNET PLUS ENVIRONMENTAuthor(s): Hua ChenJ. - Nov 2017 73 doi: 10.21506/j.ponte.2017.11.37 Abstract:With the rapid development of financial industry and internet plus technologies, more teachers need to strengthen teaching strategies methods in accounting actual behavior combination technologies. This article mainly reviews new education technologies under Internet environment on...
In recent years, the society has been developing rapidly, and it also led many new industries. The database in cloud computing is one of these At present, relational become a bottleneck computing, restricts agility WEB structure development expansibility to certain extent, so NoSQL that can break constraints RDBMS paradigm are concerned by more people. According BASE theory, easy be used highly scalable, supports massive data as main target widely field computing. This paper mainly analyzes...