SchoolWhool
Quizzes and saved progress, freeSign in freeSign in
Maths · Module 0Module material

Why Maths Matters interview questions

12 questions interviewers ask on this module’s topics, from first principles to the follow-ups. Each comes with a model answer.

  1. What does a linear model compute for one row?
  2. How do you compute predictions for a whole table, and why not a loop?
  3. A feature has the largest weight in a linear model. Is it the most important feature?
  4. Why is the mean squared error used, rather than the mean error?
  5. The mean squared error is 26,841.46. How far off is the model on a typical day?
  6. What does the derivative of the loss with respect to a weight tell you?
  7. What is the gradient, and how does gradient descent use it?
  8. How would you check that a hand-written gradient is correct?
  9. Why do probabilistic models work with sums of log-probabilities?
  10. What is underflow, and where does a float64 hit it?
  11. Two models both give the data a likelihood that prints 0.0. How do you compare them?
  12. Multiplying the day-by-day probabilities assumes something. What?