Published
Published please turn to Publication
Posts by Collection
portfolio
Uncertainties Loss
RandomCrop Uncertainties Loss combined with Dynamic Weights. The extraction of the paper with more and more varied image datasets and the largest ever scale of training, expecting a state-of-the-art performance.
Finally, we failed thought. This method was not as useful as we thought… But, it’s OK. I thought this algorithm is not mature yet. When I get more knowledge, I will revisit it.
VAE feature extractor
Developed a probabilistic generative model featuring an encoder that replaces fully connected layers with flexible neural network layers, enhancing adaptability. The decoder utilizes deconvolution to improve predictive performance and model robustness. This framework employs probabilistic embedding to capture complex data distributions, address model uncertainty, and mitigate bias, facilitating high-quality sample generation for improved downstream tasks.
publications
Enhancing Machine Learning Based SQL Injection Detection Using Contextualized Word Embedding
Published in ACMSE, 2024
Illustration of Using Machine Learning Model to Prevent SQLi Attacks
Recommended citation: https://dl.acm.org/doi/pdf/10.1145/3603287.3651187
Multi-Scale Self-Supervised Consistency Training for Trustworthy Medical Imaging Classification.
Published in In Proceedings of the 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2024
The overall architecture of the proposed multi-scale self-supervised training framework.
Recommended citation: https://ieeexplore.ieee.org/document/10782322
Improving Medical Imaging Model Calibration through Probabilistic Embedding
Published in 2024 IEEE International Conference on Big Data, 2025
Illustration of the proposed method for both training (top) and inference (bottom)
Recommended citation: https://ieeexplore.ieee.org/abstract/document/10825661/
talks
Multi-Scale Self-Supervised Consistency Training for Trustworthy Medical Imaging Classification.
Published:
