Table 12:

Validation accuracy for fully-connected neural network models with different number of layers, and different dimensions of hidden units. The dropout rate was set to 0.2 for all the models.

# of layers	Dim. of hidden units	Fold 1	Fold 2	Fold 3	Fold 4	Average
4	300	58.01	58.12	58.83	56.57	57.88
4	600	56.78	58.06	59.07	55.80	57.43
4	900	57.70	58.21	58.37	56.64	57.73
8	300	57.70	57.64	58.47	56.32	57.53
8	600	57.41	58.62	59.22	56.67	57.98
8	900	57.06	58.22	59.14	56.34	57.69
12	300	57.43	57.49	58.78	55.70	57.35
12	600	57.29	57.01	59.67	56.60	57.64
12	900	57.40	57.97	58.90	55.68	57.49