Fig. 3
From: Empirical evaluation of language modeling to ascertain cancer outcomes from clinical text reports
Impact of classification head on performance (DFCI-ImagingBERT). Associations among classification head, training dataset size, and model performance for progression/worsening (left) and response/improvement (right) for the DFCI-ImagingBERT architecture. CNN = convolutional neural network; RNN = recurrent neural network. Linear = fully connected layer applied to the BERT [CLS] token for each document