Table 1.

Hyper‐parameter space used for training the two models.

	multi‐label	single‐label
Batch size	32 or 64	32 or 64
Number of epochs	10, 15, or 20	10 or 15
Hidden size	27, 28, or 29	27, 28, 29, 210, or 211
Number of hidden layers	1, 2, or 3	1, 2, or 3
Learning rate	Between 10−5 and 5*10−3	Between 10−5 and 5*10−3
Dropout rate	Between 0 and 0.9	Between 0 and 0.9