More questions on psd and pd of matrices

In summary, PSD (Positive Semi-Definite) and PD (Positive Definite) matrices are types of square matrices with specific properties, such as all positive eigenvalues. A PD matrix is a stricter subset of PSD matrices. To determine if a matrix is PSD, you can check its eigenvalues, and for PD, you can use the Cholesky decomposition method. Non-square matrices cannot be PSD or PD. These matrices have various applications in fields like statistics, engineering, and computer science, such as optimization problems, signal processing, and machine learning algorithms. In research, they can be useful in optimization and matrix operations, data analysis, and modeling to improve accuracy and efficiency.
  • #1
insixyears
4
0
First question: If A is pd, then A^-1 is pd.
Outline of answer:
  1. If A is pd, then there exists a nonsingular matrix P st A=P'P
  2. Then A^-1 = (P'P)^-1 = (P^-1) * (P^-1)' = (P^-1)' * (P^-1)
3. If (P^-1)' * (P^-1), then there exists a A^-1 that is positive definite

Second question: If rank(Anxp) = p<n, then A'A is pd.
Outline of answer:
  1. x'(A'A)x = (Ax)'(Ax)
  2. (Ax)'(Ax) is a sum of squares which is > 0 as long as Ax does not = 0

Thanks =)

EDIT: I think I figured it out if anyone wants me to post the answers. I deleted the attempts since they look plain stupid in retrospect
 
Last edited:
Physics news on Phys.org
  • #2
crabchef said:
First question: If A is pd, then A^-1 is pd.
Outline of answer: 1. If A is pd, then there exists a nonsingular matrix P st A=P'P

Right.

2. Then A^-1 = (P'P)^-1 = (P^-1) * (P^-1)' = (P^-1)' * (P^-1)

Right. Now, denoting $Q=(P^{-1})^t$ we have $A^{-1}=Q^tQ=Q^tIQ.$

If (P^-1)' * (P^-1), then there exists a A^-1 that is positive definite

This has no sense. There exists a non singular matrix $Q$ such that $A^{-1}=Q^tIQ$ which implies $A^{-1}$ is positive definite.

---------- Post added at 09:56 AM ---------- Previous post was at 09:38 AM ----------

crabchef said:
Second question: If rank(Anxp) = p<n, then A'A is pd.
Outline of answer:1. x'(A'A)x = (Ax)'(Ax)

Right.

2. (Ax)'(Ax) is a sum of squares which is > 0 as long as Ax does not = 0

More details: denote $x=(x_1,\ldots,x_p)^t$ and $y=(y_1,\ldots,y_n)^t=Ax$, then $x^t(A^tA)x=(Ax)^t(Ax)=y^ty=y_1^2+\ldots+y_n^2\geq 0$ for all $y$. But $y_1^2+\ldots+y_n^2=0$ if and only if $y=0$ and $Ax=0$ if and only if $x=0$ (why?) which implies $A^tA$ is definite positive.
 

Related to More questions on psd and pd of matrices

1. What is the difference between PSD and PD matrices?

PSD (Positive Semi-Definite) matrices are a type of square matrix that has all positive eigenvalues, whereas PD (Positive Definite) matrices have all positive eigenvalues and are also non-singular. This means that a PD matrix is a stricter subset of PSD matrices.

2. How do I determine if a matrix is PSD or PD?

To determine if a matrix is PSD, you can check if all of its eigenvalues are positive. If they are, then the matrix is PSD. To determine if a matrix is PD, you can use the Cholesky decomposition method. If the matrix can be factored into a lower triangular matrix and its transpose, then it is PD.

3. Can a non-square matrix be PSD or PD?

No, a non-square matrix cannot be PSD or PD. PSD and PD matrices are defined as square matrices that have certain properties, such as all positive eigenvalues, which can only apply to square matrices.

4. What are some applications of PSD and PD matrices?

PSD and PD matrices have many applications in various fields such as statistics, engineering, and computer science. They are commonly used in optimization problems, signal processing, and machine learning algorithms.

5. How can I use PSD and PD matrices in my research?

If your research involves optimization or any problems that can be solved using matrix operations, then PSD and PD matrices can be useful tools. You can also use them in data analysis and modeling to improve the accuracy and efficiency of your results.

Similar threads

  • Linear and Abstract Algebra
Replies
14
Views
2K
  • Linear and Abstract Algebra
Replies
16
Views
1K
  • Linear and Abstract Algebra
Replies
4
Views
2K
Replies
7
Views
913
  • Linear and Abstract Algebra
Replies
12
Views
1K
  • Linear and Abstract Algebra
Replies
5
Views
1K
  • Linear and Abstract Algebra
Replies
3
Views
1K
  • Linear and Abstract Algebra
Replies
4
Views
1K
  • Linear and Abstract Algebra
Replies
17
Views
4K
  • Calculus and Beyond Homework Help
Replies
1
Views
530
Back
Top