Trained on the time series data instead of FC data in the first 100 WU120 subjects (no subsampling, motion censored)
