Skip to main content
. Author manuscript; available in PMC: 2016 Apr 19.
Published in final edited form as: Pac Symp Biocomput. 2016;21:144–155.

FIGURE 4.

FIGURE 4

Each panel represents one latent feature vector in the matrix Y. Vertical gray lines are manually overlaid boundaries between time bins. Time bins are ordered temporally from left to right. The weights of the original features in each latent feature representation are predominantly associated according to the time bin in which each original feature was recorded, and not by clinical similarity between the original features.