Question

We want to implement a CNN to classify fashion-MNIST images that has the following structure: ? A 2D convolution layer having 5 channels, kernel size = 3x3, padding = 2 and relu activation function ? A Max pooling layer having pool size = 2x2 and stride = 1. ? A 2D convolution layer having 15 channels, kernel size = 5x5, no padding and sigmoid activation function ? A dense layer having 100 neurons and sigmoid activation ? A dense layer having 50 neurons and relu activation ? A dense layer for the output TASK 1 Implement the CNN. After creating a random input of shape (batch_size x channel_num x height x width) = 1x1x28x28, report the output shape of all 7 layers. TASK 2 Train and test your model. Set the learning rate to 0.9 and run for 10 epochs. Report your test accuracies for batch_size = 64, 128 and 256. TASK 3 Run the LeNet-5 architecture and report the test accuracies for batch_size = 64, 128 and 256. Compare with the results in Task 2.

          We want to implement a CNN to classify fashion-MNIST images that has the following structure:
? A 2D convolution layer having 5 channels, kernel size = 3x3, padding = 2 and relu activation
function
? A Max pooling layer having pool size = 2x2 and stride = 1.
? A 2D convolution layer having 15 channels, kernel size = 5x5, no padding and sigmoid
activation function
? A dense layer having 100 neurons and sigmoid activation
? A dense layer having 50 neurons and relu activation
? A dense layer for the output
TASK 1
Implement the CNN. After creating a random input of shape (batch_size x channel_num x height x width) = 
1x1x28x28, report the output shape of all 7 layers.
TASK 2
Train and test your model. Set the learning rate to 0.9 and run for 10 epochs.
Report your test accuracies for batch_size = 64, 128 and 256.
TASK 3
Run the LeNet-5 architecture and report the test accuracies for batch_size = 64, 128 and 256.
Compare with the results in Task 2.

We want to implement a CNN to classify fashion-MNIST images that has the following structure:
? A 2D convolution layer having 5 channels, kernel size = 3x3, padding = 2 and relu activation
function
? A Max pooling layer having pool size = 2x2 and stride = 1.
? A 2D convolution layer having 15 channels, kernel size = 5x5, no padding and sigmoid
activation function
? A dense layer having 100 neurons and sigmoid activation
? A dense layer having 50 neurons and relu activation
? A dense layer for the output
TASK 1
Implement the CNN. After creating a random input of shape (batchsize x channelnum x height x width) =
1x1x28x28, report the output shape of all 7 layers.
TASK 2
Train and test your model. Set the learning rate to 0.9 and run for 10 epochs.
Report your test accuracies for batchsize = 64, 128 and 256.
TASK 3
Run the LeNet-5 architecture and report the test accuracies for batchsize = 64, 128 and 256.
Compare with the results in Task 2.

Added by Traci J.

Question

Please give Ace some feedback