Skip to main content
. 2023 Nov 28;25(1):bbad416. doi: 10.1093/bib/bbad416

Figure 2.

Figure 2

Sketch of our evaluation schemes for (A) modality imputation, (B) generation coherence, (C–E) downstream supervized tasks. Panel (C) shows the training of a classifier in the joint space of one modality, and (D) the baseline method. In panel (E), we impute the missing modality and feed the measured and the imputed profile into the baseline of panel (D).