Due Thursday, October 29: Models
“All models are wrong. Some models are useful.” George Box’s observation gives us a starting point: models simplify, and those simplifications shape what we can explain or predict. We will ask which model to use, why, and what we gain by considering several models together.
Our sentiment ensemble gives us one example. Now we connect that experience to models more broadly, including regression and classification. As you work, distinguish a model’s fit to its training data from its performance on new data.
1. The Model Thinker: prologue and chapters 1–3
Read Scott E. Page, The Model Thinker, prologue and chapters 1–3 only (about 45 book pages). Focus on why we need multiple models and the different purposes models serve.
I recommend buying the book. It is available in print and digital editions, and the photocopies are difficult to read. Scanned excerpt: prologue and chapters 1–3 (backup while you obtain a copy). Stop before chapter 4.
2. Understanding Machine Learning, chapter 1
Complete chapter 1 only of DataCamp: Understanding Machine Learning.
Chapter 2 is postponed until Tuesday, November 3.
3. Watch
- 7 Steps to Machine Learning — Google Cloud Tech (10:35).
- Fitting a Line to Data — StatQuest (9:20).