Figure 4.

t-distributed stochastic neighbor embedding (t-SNE) visualization shows discrete boundaries between poor model performance and good performance defined by Pearson correlation values. t-SNE map topographies were generated using (A) parameter combinations, (B) parameters combinations with regression coefficients for the four spindle features, and (C) spindle features. The color bar represents Pearson r values. Hyperparameters were varied until a good visualization was obtained. Perplexity was set to 30.