Table 3.

Classification accuracy for embedded group-based networks.

	Accuracy	Std Dev	Sensitivity	Std Dev	Specificity	Std Dev
	t-SNE with 99% of the variance
Younger vs. Older	88.2	5.3	85.6	11.0	89.6	7.2
	PCA with top 2 components
Younger vs. Older	88.5	6.7	81.9	13.7	92.0	8.1
	t-SNE with top 2 components
Younger vs. Older	88.9	6.6	83.1	13.7	92.0	8.1

Each measure is an average of over 100 permutations of the SVM training. The same permutations were used for all embedding methods. For both t-SNE classifications, each permutation is averaged across the 50 embedding replications as well.