Table 2 Results of the three attention mechanisms implemented as alternatives to MAGBERT. The Table includes the outcomes of the Gentamicin and the P. aeruginosa tasks as well as the average AUROC for each cohort is shown with its confidence interval (CI). “Attn” refers to the attention fusion mechanism, “Outer” refers to the tensor fusion mechanism, “MMIM” refers to Multimodal Infomax, “W.BCE” refers to weighted binary cross entropy, “C.E.” refers to cross entropy. All the Gentamicin experiments are done with a learning rate of \(5e-5\), while the P. aeruginosa ones used a learning rate of \(1e-5\).

From: Affordable and real-time antimicrobial resistance prediction from multimodal electronic health records

Data

Model

Loss

AUPRC

AUROC

\(\overline{\text {AUROC}}\)

CI Interval

Gentamicin

\(T=4, dt=1\)

LstmBertAttn

W.BCE

0.1988

0.6459

0.6198

(0.5695, 0.6701)

BertLstmAttn

W.BCE

0.0623

0.3391

LstmBertOuter

W.BCE

0.1602

0.6686

StarBertOuter

W.BCE

0.1852

0.6306

LstmBertMMIM

C.E.

0.1825

0.7565

StarBertMMIM

C.E.

0.1596

0.6782

Gentamicin

\(T=3, dt=1\)

LstmBertOuter

W.BCE

0.2209

0.5996

0.6225

(0.5699, 0.6752)

LstmBertMMIM

C.E.

0.2270

0.6555

StarBertMMIM

C.E.

0.1869

0.6125

P. aeruginosa

\(T=3, dt=1\)

LstmBertOuter

W.BCE

0.3075

0.6400

0.6650

(0.6398, 0.6902)

LstmBertMMIM

W.BCE

0.2731

0.6900