This is an old revision of the document!
====== IAM 961 HW2 ====== Use Matlab to demonstrate how uniqueness works for complex matrices, along the lines of the [[gibson:teaching:fall-2014:iam961:svddemo | SVD demo ]] in lecture. Specifically Case 1: Distinct singular values - Create a random 4 x 4 complex matrix $A$ with distinct singular values and known SVD $U_1 \Sigma_1 V_1^*$ - Compute the SVD $U_2 \Sigma_2 V_2^*$ of $A$. - Of the two SVDs, what should be the same? What is likely to be different? - Show that the third column of $V_1$ is "colinear" with the third column of $V_2$, where the constant of linearity is a complex number with unit magnitude. - Do the same for the third columns of $U_1$ and $U_2$. What is the relation between this constant and the constant of the previous question? Case 2: Repeated singular values - Create a random 4 x 4 complex matrix $A$ with $\sigma_1 = \sigma_2 > \sigma_3 > sigma_4$ and known SVD $U_1 \Sigma_1 V_1^*$ - Compute the SVD $U_2 \Sigma_2 V_2^*$ of $A$. - Of the two SVDs, what should be the same? What is likely to be different? - Show that the first two columns of $U_1$ span the same 2d subspace as the first two columns of $U_2$ (do this by showing that the first two columns of $U_1$ are in the span of the first two columns of $U_2$, and vice versa).
