Event Segmentation Update
William Villano
6/8/2020
Distribution of optimal K
This histogram depicts the distribution of optimal K (per Wasserstein Distance) for four sample subjects (across 5 runs and 6 ROIs). Event segmentation models were fit to 350 TRs of data over a range from 2 to 100 possible events:
Distribution of event lengths
This histogram depicts the distribution of event lengths for the same four sample subjects:
This histogram adds “Run #” as an additional level:
And lastly, with the added level of ROI:
Wasserstein Distance Plots
On visual inspection alone, these first three Wasserstein Distance curves below seem to show “elbows” around 20-25 events:
In contrast, this curve below does not seem to have an “elbow”:
