Machine Learning · Module 3Module material
Linear Regression and Gradient Descent interview questions
15 questions interviewers ask on this module’s topics, from first principles to the follow-ups. Each comes with a model answer.
Linear regression
- How is a simple linear regression line fitted, and what do its two numbers mean?●
- Why squared error and not the sum of residuals or absolute error?●
- What is R², and what is it not?●
- What does a coefficient mean in multiple regression, and why did weight's change from −7.80 to −6.85 when model year was added?●●
- How do you compare the importance of two features in a linear model?●●
Gradient descent
- What is gradient descent, and what does the learning rate do?●
- Batch, stochastic and mini-batch gradient descent: what is the difference?●●
- Does scikit-learn's
LinearRegressionuse gradient descent?●● - Why does feature scaling speed up gradient descent?●●
Collinearity and diagnostics
- What is multicollinearity, and what does it do to a model?●
- What is the variance inflation factor, and how do you use it?●●
- What is the dummy-variable trap?●
- What do you look for in a residual plot?●●
- Outlier, leverage, influence: what is the difference?●●
- When would you use a robust loss such as Huber, and when not?●●