Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Hello all,

This is a question that might or might not be related to my previous one.

Imagine you have two matrices:

  1. Matrix $\mathbf{\Phi}=[\Phi_1,\ldots,\Phi_M]\in\mathbb{R}^{L\times M}$ where $M\leq L$. The column-vectors have unit norm $||\Phi_i||=1$, and the maximum dot-product between distinc vectors is $|\Phi^\top_i\Phi_j|\leq\alpha$ for any $i\neq j$.

  2. The second matrix $\mathbf{T}\in\mathbb{R}^{L\times L}$ is totally known. It is positive-definite and its eigenvalues have multiplicity 1, i.e. $\lambda_1 > \ldots > \lambda_L > 0$.

My question is... what are all the possible diagonal values of the matrix product $\mathbf{\Phi}^\top\mathbf{T}\mathbf{\Phi}$?

I actually know the solution by the case $\alpha=0$:

In such case $\mathbf{\Phi}$ is simply an orthogonal matrix. Call $\mathbf{V}$ and $\mathbf{\Lambda}$ the eigenvector and eigenvalue matrix of $\mathbf{T}$ respectively. Thus $\mathbf{\Phi}^\top\mathbf{T}\mathbf{\Phi} = \left(\mathbf{\Phi}^\top\mathbf{V}\right)\mathbf{\Lambda}\left(\mathbf{V}^\top\mathbf{\Phi}\right)$ where $\mathbf{\Phi}^\top\mathbf{V}$ is another "orthogonal" matrix (rows are orthogonal). Thus the original question becomes "what diagonal values are possible with a matrix with prescribed eigenvalues?"

This is completely answered by the Schur-Horn theorem and relies on some majortization theory.

Thank you!

share|improve this question
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.