Question: (SVD of score matrix) An exam with m questions is given to n students. The instructor collects all the grades in a n × m matrix G, with Gij the grade obtained by student i on question j. We would like to assign a difficulty score to each question, based on the available data.
1. Assume that the grade matrix G is well approximated by a rankone matrix sq?, with s ∈ Rn and q ∈ Rm (you may assume that both s, q have non-negative components). Explain how to use the approximation to assign a difficulty level to each question. What is the interpretation of vector s?
2. How would you compute a rank-one approximation to G? State precisely your answer in terms of the SVD of G.