DNN Prediction
Deep neural network (DNN) prediction focuses on improving the accuracy, reliability, and interpretability of DNN predictions across diverse applications. Current research emphasizes enhancing robustness through techniques like ensemble averaging and control-theoretic online adaptation, as well as quantifying prediction uncertainty using various methods categorized by uncertainty source (data vs. model). These advancements are crucial for deploying DNNs in high-stakes domains requiring trustworthy predictions and for gaining a better understanding of their decision-making processes, thereby increasing their reliability and acceptance in scientific and practical settings.
Papers
October 15, 2024
February 1, 2024
February 26, 2023
January 17, 2023
December 14, 2022
June 17, 2022
March 7, 2022