A5027810378- Text and Document Classification Technologies
- Machine Learning and Data Classification
- Topic Modeling
- Advanced Computational Techniques and Applications
- VLSI and Analog Circuit Testing
- Anomaly Detection Techniques and Applications
- Natural Language Processing Techniques
- Domain Adaptation and Few-Shot Learning
- Advanced Image Fusion Techniques
- Energy Load and Power Forecasting
- Advanced MRI Techniques and Applications
- Advanced Neural Network Applications
- Time Series Analysis and Forecasting
- Software Testing and Debugging Techniques
- Fault Detection and Control Systems
- Advanced Image Processing Techniques
- Advanced Text Analysis Techniques
- Sentiment Analysis and Opinion Mining
- Image and Signal Denoising Methods
Yunnan Normal University
2022-2024
Abstract Image denoising is one of the hottest topics in image restoration area, it has achieved great progress both terms quantity and quality recent years, especially after wide intensive application deep neural networks. In many learning based models, performance can greatly benefit from prepared clean/noisy pairs used for model training, however, also limits these models real scenes. Therefore, more researchers tend to develop that be learned without pairs, namely well generalised...
Test-time augmentation (TTA) has become a widely adopted technique in the computer vision field, which can improve prediction performance of models by aggregating predictions multiple augmented test samples without additional training or hyperparameter tuning. While previous research demonstrated effectiveness TTA visual tasks, its application natural language processing (NLP) tasks remains challenging due to complexities such as varying text lengths, discretization word elements, and...
Mixup is an effective data augmentation method that generates new augmented samples by aggregating linear combinations of different original samples. However, if there are noises or aberrant features in the samples, mixup may propagate them to leading over-sensitivity model these outliers. To solve this problem, paper proposes a called AMPLIFY. This uses attention mechanism Transformer itself reduce influence and values on prediction results, without increasing additional trainable...
Human beings have rich emotions, in which positive emotions need to be constantly maintained, while negative regulated since most mental diseases are caused by the long-term persistence of emotions. Aiming at emotion classification sentiment analysis tasks, a deep residual BiGRU (Bidirectional Gated Recurrent Unit) neural network model is introduced improve effect and solve problem long-distance dependence between layers RNN (Recurrent Neural Network) using recurrent correlation channel...
Mixup is an effective data augmentation method that generates new augmented samples by aggregating linear combinations of different original samples. However, if there are noises or aberrant features in the samples, may propagate them to leading over-sensitivity model these outliers . To solve this problem, paper proposes a called AMPLIFY. This uses Attention mechanism Transformer itself reduce influence and values on prediction results, without increasing additional trainable parameters,...
Test-time augmentation (TTA) is a well-established technique that involves aggregating transformed examples of test inputs during the inference stage. The goal to enhance model performance and reduce uncertainty predictions. Despite its advantages not requiring additional training or hyperparameter tuning, being applicable any existing model, TTA still in early stages field NLP. This partly due difficulty discerning contribution different samples, which can negatively impact In order address...
Test-time augmentation (TTA) is a technique that improves model predictive performance by aggregating augmented samples without requiring additional training. While TTA works with any data modality, its use in natural language processing (NLP) limited. This partly because the text complex and inappropriately introducing noise can easily change semantics syntax, making it difficult to ensure label consistency. Existing strategies lack generality are only effective for certain datasets,...