Fig. 4From: Trade-off between conservation of biological variation and batch effect removal in deep generative modeling for single-cell transcriptomicsThe trade-off curves of surrogate metrics. Compare Pareto MTL with MINE and Pareto MTL with MMD to estimate three types of trade-offs of surrogate metrics for all Pareto candidates and non-dominated points on the TM-MARROW dataset: A, B trade-offs between generative loss \(\overline{U}_n\) and mutual information estimator NN; C, D the trade-offs between negative ASW and negative BE; E, F the trade-offs between negative NMI and negative BE. The point with the largest marker size in C and E corresponds to scVI alone (i.e. minimizing \(U_n\))Back to article page