You are given a very large and sparse numeric ratings matrix M whose rows represent patrons in a cafe-review network. columns represent specific cafes. The ratings in the matrix (where present) are whole numbers in the range of 1 to 5. There are a lot of missing entries.

Let the complete matrix (with no missing information) be M1 and M2 be the matrix obtained by centering M1, i.e. subtracting the means from each row. If M2 is decomposed using Singular Value Decomposition (SVD) what information is contained in the factors obtained by the decomposition?