Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm trying to expand this Frobenius form $||C \circ (A-XB)||_F^2$ (here $\circ$ is the Hadamard point-wise multiplication). I want to find the minimum value with respect to X.

$$ \frac{\partial}{\partial X}||C \circ (A-XB)||_F^2 = 0$$

I've being trying to develop using the fact that the Frobenius form is $||A||_F^2=trace(AA^*)$ but the Hadamard product is always on my way.

How would you approach this?

share|cite|improve this question
    
You can put $g(X)=C\circ (A-XB)$ where $\circ$ is the Hadamard product, and use the chain rule. – Davide Giraudo Jan 30 '12 at 16:41

You need to use the chain rule. $$\partial (\mathbf{X} \circ\mathbf{Y}) = \partial \mathbf{X} \circ \mathbf{Y} + \mathbf{X} \circ \partial \mathbf{Y} $$

Also, the fact that $$||\mathbf{A}||_2^2 = trace(\mathbf{A}^T\mathbf{A})$$

Look the following reference for matrix calculus it has all the stuff you would need.

Matrix Cookbook: [http://orion.uwaterloo.ca/~hwolkowi/matrixcookbook.pdf][1]

share|cite|improve this answer

Given $f = \|M\|_F^2 = M:M$, the differential is $\,\,df=2M:dM$

Substituting $M=C\circ(A-XB)$, we have $$ \eqalign { df &= 2(C\circ(A-XB)) : d(C\circ (A-XB)) \cr &= -2(C\circ(A-XB)) : C\circ dX\,B \cr &= -2(C\circ(A-XB)\circ C) : dX\,B \cr &= -2(C\circ(A-XB)\circ C)B^T : dX \cr &= \bigg(\frac {\partial f} {\partial X}\bigg) : dX \cr } $$ Setting the derivative to zero yields $$ \eqalign { \frac {\partial f} {\partial X} &= 0 = -2\,(C\circ(A-XB)\circ C)B^T \cr (XB\circ C\circ C)B^T&= (A\circ C\circ C)B^T \cr } $$ Now what is the best way to solve for $X$?

share|cite|improve this answer
    
Hi greg, Did you known how the solve this for X? I have the same question about this. math.stackexchange.com/questions/1670377/… Thanks. – BioChemoinformatics Feb 28 at 23:02

Your Answer

 
discard

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.