(1) Use this walkthrough to setup Python with Keras in Google Drive: https://medium.com/deep-learning-turkey/google-colab-free-gpu-tutorial-e113627b9f5d (You don't have to read the whole thing---you might find it helpful/interesting to do so, but for now you only need to get up to the point where tensorflow 2.0 and keras are installed and imported.) (2) Put the attached two Jupyter notebook files in your Google drive colab folder and read through them line-by-line and try to understand what each command does. (There's no comments, so you might have to look up some of the commands, but most of them are self-explanatory so I think you'll be able to figure it out just by looking at the code.) If you're stuck/confused by any parts of it, ask me by email (I'll also quickly walk through it next week at the beginning of class). (3) Play around with the hyperparameters and neural net architecture (feel free to be creative and experiment and try different things even if you don't know how well they'll do!) and see what is the best validation accuracy you can acheive for a MLP network (fully connected layers of hidden neurons, the kind we've been studying before) and for a CNN network (the convolution kind we just started studying). HINT: you don't need to use all the code in these files for this, only the part up to the "history" where the model is trained and then the following command that computes the validation accuracy (all the code after that is just to randomly sample an image in the database and see how the model labels it). You don't have to turn anything in here, but we'll start class next week with a fun informal competition to see which student in the class got the highest accuracies for the two types (MLP and CNN) and then have the student(s) either describe their choices or share their screen and show the rest of us.