Fig. 8
From: Conditional similarity triplets enable covariate-informed representations of single-cell data
Visualizing Samples in 2d Based on Learned Embeddings. PCA was used to project samples in two-dimensions with PCA, based on the learned embedding coordinates under CytoSet (top) and CytoCoSet (bottom). Thirty classification trials were run and we visualized projections of test set samples corresponding to the trial achieving the median AUC (denoted AUCs are those obtained with CytoCoSet). Dots represent samples and are colored by the class label. Gray dotted lines represent a possible decision boundary that could be used to separate the samples correctly