Multitask Balanced and Recalibrated Network for Medical Code Prediction
Published in Preprint, 2021
Recommended citation: Sun, Wei, et al. "Multitask Balanced and Recalibrated Network for Medical Code Prediction." arXiv preprint arXiv:2109.02418 (2021). https://arxiv.org/pdf/2109.02418.pdf
We extend our previous work~(MT-RAM) to a more advanced model called~(MARN). MARN leverages focal loss to deal with the imbalanced class problem of MIMIC-III dataset. Multitask learning scheme can capture code associations by joint-training two coding branches. RAM is designed to refine textual features extracted from lengthy and noisy clinical notes.
Recommended citation: Sun, Wei, et al. “Multitask Balanced and Recalibrated Network for Medical Code Prediction.” arXiv preprint arXiv:2109.02418 (2021).