Table 2:

Trial Numbers for Figure 2, Comparing Nonlinear Activation Functions for the Dendrite Nodes in the k-tree For Each Dataset.

MNIST
	1-tree	2-tree	4-tree	8-tree	16-tree	32-tree
ReLU	8	9	10	9	10	10
LReLU	8	10	10	10	10	10
Sigmoid	1	5	0	2	0	0
Swish	10	10	10	10	10	10
NaCaK	10	10	10	10	10	10
FMNIST
	1-tree	2-tree	4-tree	8-tree	16-tree	32-tree
ReLU	7	10	8	10	10	10
LReLU	7	10	10	10	10	10
Sigmoid	2	3	0	0	0	1
Swish	4	7	8	9	10	10
NaCaK	9	10	10	10	10	10
KMNIST
	1-tree	2-tree	4-tree	8-tree	16-tree	32-tree
ReLU	8	9	9	10	10	10
LReLU	7	7	10	10	10	10
Sigmoid	1	1	1	0	0	0
Swish	10	10	10	10	10	10
NaCaK	9	10	10	10	10	10
EMNIST
	1-tree	2-tree	4-tree	8-tree	16-tree	32-tree
ReLU	10	10	10	10	10	10
LReLU	10	10	10	10	10	10
Sigmoid	8	3	1	2	0	0
Swish	10	10	10	10	10	10
NaCaK	10	10	10	10	10	10
SVHN
	1-tree	2-tree	4-tree	8-tree	16-tree	32-tree
ReLU	0	0	0	1	0	1
LReLU	1	1	1	1	1	4
Sigmoid	0	0	0	0	0	0
Swish	4	2	2	3	3	3
NaCaK	2	3	4	4	3	2
USPS
	1-tree	2-tree	4-tree	8-tree	16-tree	32-tree
ReLU	6	8	10	10	10	10
LReLU	6	9	10	10	10	10
Sigmoid	3	3	2	0	0	0
Swish	10	10	10	10	10	10
NaCaK	8	6	7	10	10	10
CIFAR10
	1-tree	2-tree	4-tree	8-tree	16-tree	32-tree
ReLU	2	4	4	5	5	6
LReLU	2	1	5	3	7	6
Sigmoid	0	0	0	0	0	0
Swish	1	4	3	2	4	6
NaCaK	10	9	8	8	9	8