A5101521826- Anomaly Detection Techniques and Applications
- Computational Drug Discovery Methods
- Music and Audio Processing
- Time Series Analysis and Forecasting
- Civil and Geotechnical Engineering Research
- Machine Learning and Algorithms
- Simulation and Modeling Applications
- Geophysical and Geoelectrical Methods
- Peer-to-Peer Network Technologies
- Interactive and Immersive Displays
- Neural Networks and Applications
- Hand Gesture Recognition Systems
- Caching and Content Delivery
- Pharmacological Effects of Natural Compounds
- Water Systems and Optimization
- Protein Structure and Dynamics
- Gene expression and cancer classification
- Speech and Audio Processing
- Tactile and Sensory Interactions
- Face and Expression Recognition
- Advanced Computational Techniques and Applications
- Control Systems and Identification
- Recommender Systems and Techniques
- Analytical Chemistry and Chromatography
- Plant-based Medicinal Research
Tsinghua University
2022-2023
Shanghai Jiao Tong University
2019-2022
China University of Petroleum, East China
2020-2021
Chinese Academy of Sciences
2021
Institute of Geology and Geophysics
2021
University of Chinese Academy of Sciences
2021
Beijing University of Posts and Telecommunications
2005-2006
Automatic detection of machine anomaly remains challenging for learning. We believe the capability generative adversarial network (GAN) suits need audio detection, yet rarely has this been investigated by previous work. In paper, we propose AEGAN-AD, a totally unsupervised approach in which generator (also an autoencoder) is trained to reconstruct input spectrograms. It pointed out that denoising nature reconstruction deprecates its capacity. Thus, discriminator redesigned aid during both...
Background: Drug development requires a lot of money and time, the outcome challenge is unknown. So, there an urgent need for researchers to find new approach that can reduce costs. Therefore, identification drug-target interactions (DTIs) has been critical step in early stages drug discovery. These computational methods aim narrow search space novel DTIs elucidate functional background drugs. Most developed so far use binary classification predict presence or absence between target....
Magnetic anomaly detection (MAD) is used for detecting moving ferromagnetic targets. In this study, we present an end-to-end deep-learning model magnetic on data recorded by a single static three-axis magnetometer. We incorporate attention mechanism into our network to improve the capability of long time-series signals. Our has good performance under Gaussian colored noise with power spectral density 1/fα, which similar field noise. method does not require another magnetometer eliminate...
Traditional Chinese medicine has been used to treat and prevent infectious diseases for thousands of years, accumulated a large number effective prescriptions. Deep learning methods provide powerful applications in calculating interactions between drugs targets. In this study, we try use the method deep reposition molecules medicines (CMs) targets syndrome coronavirus 2 (SARS-CoV-2). A convolution neural network with residual module (DCNN-Res) is constructed trained on KIBA dataset. The...
Recently, cancer classification based on gene expression has been developed. This gives a hope for the discrimination of to more systematic direction. However, there're many challenges existing in new method. Maybe most important one is unbalance that so few training samples exist compared huge genes collected. So feature selection becomes center problem classification. A novel mixture strategy proposed this paper, it make use characters filter and wrapper, synthesis three methods: Pearson...
The content-based publish/subscribe system is an effective paradigm for implementing on-demand event distribution. Each needs to be matched against subscriptions identify the target subscribers. To improve matching performance, many novel data structures have been proposed. However, predicates contained in are handled same way most existing structures, without considering their differences probability. In this paper, we propose concept of parallel ensemble (PEM) based on subscription...
Pretrained deep belief networks have been shown to speed up the convergence of models and improve their performance on many supervised recognition tasks. However, its unsupervised anomalous sound detection remains be explored. In this paper, we initialize parameters autoencoder (AE) variant with pretrained network (DBN) use them for detection. We explore effect number layers initialized pre-training DBN different data performance. Experimental results show that appropriate can substantially...
Learning theory based on ERM principle, especially promoted by VC provides some conditions the hypothesis space to ensure generalization. However, several successful learning algorithms including regularization learning, SVM, bagging and boost are not strictly ERM. So, scientists looking for new foundation of learning. Stability perhaps foundation. We give an exponential bound generalization performance concentration inequality with strong CV stability.
Smart wearable devices have become pervasive as they are portable and intelligent. The popular method to interact with it is touch-screen, which error-prone cumbersome due its limited size. There a few innovative works designing virtual dial plate on the hand back, need special-purpose sensors or microphones may suffer from privacy leak. We propose MType, system only employs already built in commercial-off-the-shelf (COTS) device magnetic ring expand interaction space between users devices....