Description
Questions:
1. [20 marks] Implement an edge filter to filter an image. The convolution should be implemented with
basic array processing without the use of a convolution function or an exciting edge filter. Provide
your answer as Jupyter file.
2. [80 marks] Go through the Tensorflow tutorials, specifically
https://www.tensorflow.org/get_started/get_started
and
https://www.tensorflow.org/get_started/mnist/beginners
and
https://www.tensorflow.org/get_started/mnist/pros
Then answer the following questions. Provide your answers as Jupyter files (.ipynb).
2.1. [20 marks, 10 marks for Grads] Get familiar with the code provided in the tutorial with 2
convolutional layers and save it in MNIST_conv2.ipynb. Try to run this code. If the code is running
into an error message, try to track the error message. You may need to use batches to test the model. In
the provided code, they test the whole data set at once. If your system has not enough memory, you can
use batches of the test data (check how it is done in the training phase).
2.2. [20 marks, 10 marks for Grads] Modify the code to have only one convolutional layer by removing
the second convolutional layer. Save your program in the MNIST_conv1.ipynb.
2.3. [20 marks, 10 marks for Grads] Modify the code MNIST_conv2.ipynb to have 3 convolutional
layers. Add another convolutional layer with filter size (5 by 5) after the second convolutional layer.
Save this code in MNIST_conv3.ipynb.
2.4. [20 marks, 10 marks for Grads] Evaluate the performance of networks with 1, 2 or 3 layers and
report on your findings.
2.5. Grad Students only [20 marks] Add noise to the training data set and train the two-layer model. To
make mislabeled data you can simply shuffle some of the labels. For example, if you want to make 10
percent of the data mislabeled, you can permute/shuffle %10 of the labels.
2.6. Grad Students only [20 marks] Evaluate the performance of the network with different rates of noise
(%10, %25, %50, %75, %100) applied. Illustrate your result in a plot. Explain the performance of your
network with %100 noise, specifically the difference between the training and test error.
