Due: March 26, 2025

Purpose: Working with the MNIST dataset containing hand-written digits, the goal is to create a model, train it on the dataset, and see how well it learns.

Reading: Chollet, Chapter 1 and Chapter 2, Section 2.1

Setup. The document Initial Setup describes how to setup Anaconda and then how to set up a virtual environment for TensorFlow/Keras (and other things).

• We've already set up Anaconda, so there's no reason to repeat the first section, so skip First time initialization for Anaconda (though if you don't want Anaconda permanently enabled when you start Terminal, have look at the part where I describe how to turn it off)
• Run through the procudure for Installing Keras
• Look at the procedures in Installing PyTorch. While we don't need PyTorch for this homework, we may need it later. Installation is optional (for now)
• We've already done the section Starting out with matplotlib
• The three links at the bottom of the page are useful for improving the look of your Markdown cells
  

Assignment. Work through the example code in Section 2.1, executing each code listing in the text (starting with Listing 2.1 and working through Listing 2.7) in its own code block in a Jupyter Notebook. Be sure to give your notebook page an appropriate title, with your name somewhere near the top. Be careful about the following:

• Run each block of code individually as you enter them. Do not try to run the entire notebook at once, because if you have a typo, it will be very diffult to track down!
   
  
• Be aware that some of the code blocks in the text show interaction with the Python interpreter, which you should NOT copy into the notebook. For example, the two blocks after Listing 2.1 look like this:

  	    >>> train_images.shape
  	    (60000, 28, 28)
              >>> len(train_labels)
  	    60000 
              >>> train_labels
              array([5, 0, 4, ..., 5, 6, 8], dtype=uint8)

  Recall that the >>> is the Python prompt. Just enter and run each command in its own code block:

  	    train_images.shape
  	    
              len(train_labels)
  	    
              train_labels

   You should get the same results as above.
   
  
• Make a note of any Python code that you aren't certain about and ask in class. I want everyone to have a good handle on the Python. Of course, much of the code is made up of Keras functions that we don't know yet...they'll be discussed in detail later in the text.
   
  
• Everything worked properly if the last code block (from Listing 2.7) returns a test accuracy of 0.97 or better. Even if your accuracy isn't quite that good, looking at what worked differently will be illumnating - don't think you have to run everything again.  
  

Deliverables

When complete, select File -> Download As from the menu and choose to download as HTML. Save the HTML version of your notebook in your ~/html/CS356 directory and send me a link.

---