Scribd
Upload
20 views1 page

LeNet5 Assignment Solved

The document outlines the architecture of the LeNet-5 neural network, detailing each layer's parameters including kernel size, stride, zero padding, output feature map size, number of trainable parameters, and activation functions. It includes calculations for the number of parameters and connections for each layer, culminating in the final output layer representing digits 0-9. The table provides a comprehensive overview of the model's structure and its computational requirements.

Uploaded by

kimoachaia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
20 views1 page

LeNet5 Assignment Solved

The document outlines the architecture of the LeNet-5 neural network, detailing each layer's parameters including kernel size, stride, zero padding, output feature map size, number of trainable parameters, and activation functions. It includes calculations for the number of parameters and connections for each layer, culminating in the final output layer representing digits 0-9. The table provides a comprehensive overview of the model's structure and its computational requirements.

Uploaded by

kimoachaia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

LeNet-5

Find the missing in the following table:

Layer Kernel Stride Zero Size of # trainable # of Activation


Size Padding output parameters connections function
feature map
or image
Input 32 x 32 x 3
image
C1 5x5 1 No 6 @ 28 x 28 5x5x6 156 × 28 × Tanh
weights + 6 28 =
offsets = 122,304
150
weights + 6
offsets =
156
S2 2x2 2 No 6 @ 14 x 14 6 Weights (2×2 + 1) × Sigmoid
(average) + 6 × 14 × 14
6 offsets = = 5,880
12
C3 5x5 1 No 16 @ 10 x 10 6 × (3×5×5 10 × 10 × Tanh
+ 1) + 6 × 1,516 =
(4×5×5 + 1) 151,600
+3×
(4×5×5 + 1)
+1×
(6×5×5 + 1)
= 1,516
S4 2×2 2 No 16 @ 5 x 5 2 × 16 = 32 (2×2 + 1) × Sigmoid
16 × 5 × 5 =
2,000
C5 5x5 1 No 120 @ 1 x 1 120 × 48,120 Tanh
(16×5×5 +
1) = 48,120
F6 - - - 84 neurons 84 × (120 + - Tanh
1) = 10,164
F7 - - - 10 neurons 84 × 10 = - Softmax
(representing 840
digits 0–9)

You might also like