Project 2 for the ECE188 course Spring 22.
This project has two parts:
- Training an MNIST network.
- Producing and exploring adversarial examples in Neural Nets.
The Colab notebooks has all the information required for the project.
While Task 1 can be run directly, here are instructions for running Task 2 and 3:
- Add the images in the
imagesdirectory to your Google Drive. The images are taken directly from the Imagenet database. - Update the path for your images in Google Drive before running the code
Note: Task 3 may take as long as 5 to 10 minutes to run. This is due to the intensive computations involved in the adversarial attack.
Submission:
-
All your code, tables and text need to be contained in the jupyter notebooks. Hence, your final submission would contain 2 notebooks, one for each part.
-
Also upload the two notebooks to a public github repo and attack the link of the github repo at the top of one of your notebooks.
DUE DATA : 10th June 2022