Skip to main content
. 2018 May 22;18(5):1654. doi: 10.3390/s18051654

Table 2.

Average AUC (standard deviation) and corresponding average training time per neural network (NN) architecture type for a subset of the data. The difference in training time between the ways of splitting the data is due to the slower convergence when splitting at the sample level.

Subject Level Sample Level
AUC Time (h) AUC Time (h)
CNN 0.52 (0.07) 6 0.74 (0.07) 7
LSTM 0.61 (0.10) 160 0.91 (0.06) 180
ConvLSTM 0.60 (0.09) 35 0.90 (0.05) 40