Week 11: Journal Questions
Answer the following question for
Monday's reading about designing ML systems (R&N Section 19.9)
- What does it mean when we say our data has "unbalanced classes"?
- What problems can be caused by imbalanced data, and how can we address these problems?
Answer the following question for
Wednesday's reading from the Lecture Notes on Measuring Performance (Chapter 10, but focusing on 10.1 for these questions)
- Why is accuracy not always sufficient as a measure of a classifier's performance?
- What do the entries in a Confusion Matrix represent?
Answer the following question for
Friday's reading about evaluating what machines have learned (Mitchell Chapter 6)
-
In this chapter, Mitchel asks "are we fooling ourselves when
we think these networks have actually learned concepts that we are
trying to teach them?" Cite at least two examples mentioned in the
reading that prompted this question.
-
Based on what you read, are you more or less likely to trust an autonomously driven car? Briefly explain your answer.