Skip to main content
. Author manuscript; available in PMC: 2022 Dec 1.
Published in final edited form as: Proc IEEE Int Conf Data Min. 2021 Dec;2021:857–866. doi: 10.1109/icdm51629.2021.00097

TABLE V.

Prediction results of 25 Phenotypes by benchmarking LSTM [5] model under different losses. BCE: Multi-label Binary Cross Entropy; CBCE: Multi-label Contrastive Binary Cross Entropy; CSCE: Multi-label Contrastive Softmax Cross Entropy; SCR: Multi-label Supervised Contrastive Regularizer. We highlight the best performance w.r.t different metrics.

Micro AUROC Macro AUROC Weighted AUROC

BCE 0.822 0.772 0.758
BCE+λSCR 0.824 0.775 0.761
CBCE+λSCR 0.823 0.774 0.761
CSCE+λSCR 0.824 0.774 0.761