Due: April 02, 2025

Purpose: complete our comparison of the simple DLN models from Chapter 4, which sets us up for the coming chapters. This assignment is worth 50 points.

Reading:

• Chollet: Chapter 3; Chapter 4, Section 4.3
  

Problems.

1. (12 points) The final example in Chapter 4 shows how to do regression, or predicting the value of a continuous variable (the classification examples we'll do in class this week all output discrete values). In this case, the DLN will attempt to predict the median price of a single-family dwelling in California. This will include reading Sections 4.7 through 4.12 working through Listings 4.24 - 4.33 including the text boxes showing predictions, etc.
   
   Build and run this model in a Jupyter notebook, and post your results as we have in the Deep Learning exercises.
   
   
2. (10 points) In a NEW notebook (not the one you just did), use Markdown cells to document the same sort of comparison we did in the lab, comparing the regression model with the NewsWire model this time. Again this is similarities and differences in the two models with respect to:
   
   Number and type of layers
   Number and type of inputs
   Number and type of output
   Activation, optimizer, and loss functions
   Resulting Training/Validation accuracy
   
   
    
   
3. (12 points)  Find a paper or article online discussing either the ChatGPT bot, or the GPT-3 model it is based on. The article should be moderately technical, not written for a non-tech audience. Also do not choose an article that overhypes the technology. You want a discussion of the underlying machine learning concepts.
   
   Summarize the specific machine learning concepts used, and provide title, authors, and a link to the article. In particular try to find out the size and scope of the training data. Where did it come from?
    
   
4. (16 points) Read the paper Deep Learning by Hinton/LeCun/Bengio (collectively known as the godfathers of deep learning); you should have access through your library account. This paper serves as an overview of the entire field circa 2016 and intro to this part of this course.
    
   Write a summary of the main ideas in as many paragraphs as you need; focus on the following topics:
   
   • Supervised Learning
   • The backprop algorithm
   • CNNs for image recognition (and other tasks)
   • RNNs for language processing (and other tasks)
     

Deliverables

• Add HTML versions of your Jupyter notebooks either on the CS webserver (your ~/html/CS356 folder) or post directly to BrightSpace
• Post your article summaries (in PDF format if possible) to BrightSpace
  

---