Бірлескен шығын функциясы
method
ID: jointlossfunction
Related Papers
Doubly-Trained Adversarial Data Augmentation for Neural Machine Translation
AbstractNeural Machine Translation (NMT) models are known to suffer from noisy inputs. To make models robust, we generate adversarial augmentation samples that attack the model and preserve the source-side meaning …...
Patient Outcome and Zero-shot Diagnosis Prediction with Hypernetwork-guided Multitask Learning
AbstractMultitask deep learning has been applied to patient outcome prediction from text, taking clinical notes as input and training deep neural networks with a joint loss function of multiple tasks. …...