This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
Uncertainty-Aware Generative Oversampling Using an Entropy-Guided Conditional Variational Autoencoder
Created by
Haebom
Author
Amirhossein Zare (SeyedAbolfazl), Amirhessam Zare (SeyedAbolfazl), Parmida Sadat Pezeshki (SeyedAbolfazl), Herlock (SeyedAbolfazl), Rahimi, Ali Ebrahimi, Ignacio Vazquez -Garcia , Leo Anthony Celi
Outline
LEO-CVAE is a novel generative oversampling framework for solving the imbalanced learning problem of high-dimensional biomedical data. Based on CVAE, it explicitly incorporates local uncertainty into the learning and data generation processes. It utilizes local entropy as an uncertainty measure, enabling more robust learning in regions of high uncertainty and focusing synthetic sample generation on these regions. It was applied to clinical genomic datasets (ADNI and TCGA lung cancer) and demonstrated improved classification performance compared to existing oversampling and generative models.
Takeaways, Limitations
•
Takeaways:
◦
Demonstrating the effectiveness of generative oversampling techniques using uncertainty information.
◦
Suggesting the possibility of solving imbalanced learning problems for data with complex nonlinear structures, such as biomedical data.
◦
Achieves superior performance compared to traditional oversampling methods and existing generative models.
•
Limitations:
◦
Lack of information on specific LEO-CVAE implementation methods, hyperparameter settings, etc.
◦
Generalization performance needs to be verified to other domains and datasets.
◦
Further research is needed to optimize the calculation and utilization of local entropy.
