Activations:
Activation Gradients:
Weights:
(
)(
)(
)(
)(
)(
)(
)(
)
Weight Gradients:
(
)(
)(
)(
)(
)(
)(
)(
)
conv (24x24x8)
filter size 5x5x1, stride 1
max activation: 2.38866, min: -3.0595
max gradient: 0.00006, min:
-0.00009
parameters: 8x5x5x1+8 = 208
Activations:
Activation Gradients:
relu (24x24x8)
max activation: 2.38866, min: 0
max gradient: 0.00008, min: -0.00009
Activations:
Activation Gradients:
pool (12x12x8)
pooling size 2x2, stride 2
max activation: 2.38866, min: 0
max gradient: 0.00008, min:
-0.00009
Activations:
Activation Gradients:
Weights:
(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)
Weight Gradients:
(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)(
)
conv (12x12x16)
filter size 5x5x8, stride 1
max activation: 6.37392, min: -5.23381
max gradient: 0.00005, min:
-0.0001
parameters: 16x5x5x8+16 = 3216
Activations:
Activation Gradients:
relu (12x12x16)
max activation: 6.37392, min: 0
max gradient: 0.00007, min: -0.0001
Activations:
Activation Gradients:
pool (4x4x16)
pooling size 3x3, stride 3
max activation: 6.37392, min: 0
max gradient: 0.00007, min: -0.0001
Activations:
Activation Gradients:
fc (1x1x10)
max activation: 8.75213, min: -6.32776
max gradient: 0.00022, min: -0.00028
parameters: 10x256+10
= 2570
Activations:
softmax (1x1x10)
max activation: 0.99972, min: 0
max gradient: 0, min: 0