IITM MLT Week 5 Questions
Linear and Kernel Regression
Week 5 question PDF
Practice fitting regression models and evaluating their predictions. Numerical datasets, optimization questions and kernel formulations connect squared-error loss with the weights learned from training data.
The complete question statements, options, numerical data and any supporting graphs or tables are in the PDF. Numbered comprehension questions can include several subquestions. Answer keys are included for checking your attempt; these are not full worked solutions.
Topics covered in this collection
- Least squares and linear regression
- Gradient descent and stochastic gradient descent
- Kernel regression
- Prediction error and validation
How to approach these questions
- Check the orientation of the design matrix before applying a closed-form least-squares expression.
- State whether the loss is a sum or a mean before calculating gradients or mean squared error.
- Keep training, validation and held-out predictions separate when comparing regression models.
Attempt the questions using the assumptions and notation stated in each problem. For short-answer questions, round only at the final step to the requested precision. Check the provided key after solving and revisit the relevant concept when an answer differs.