If A is nnd, then show that I + A is pd

  • Context: MHB 
  • Thread starter Thread starter insixyears
  • Start date Start date
  • Tags Tags
    Pd
Click For Summary
SUMMARY

The discussion centers on the relationship between non-negative definite (nnd) matrices and positive definite (pd) matrices, specifically addressing the assertion that adding the identity matrix to an nnd matrix results in a pd matrix. It is established that for a matrix A to be nnd, the condition $x'Ax \geq 0$ must hold for every nonzero vector x. The addition of the identity matrix I ensures that $x'(I+A)x > 0$, confirming that I + A is indeed pd. The conversation clarifies misconceptions regarding the implications of matrix entries on definiteness.

PREREQUISITES
  • Understanding of matrix definitions: non-negative definite (nnd) and positive definite (pd)
  • Familiarity with quadratic forms and the expression $x'Ax$
  • Knowledge of matrix operations, particularly addition and multiplication
  • Basic linear algebra concepts, including vector spaces and eigenvalues
NEXT STEPS
  • Study the properties of positive definite matrices and their applications in optimization
  • Learn about eigenvalues and eigenvectors in the context of matrix definiteness
  • Explore the implications of matrix definiteness in machine learning algorithms
  • Investigate the Cholesky decomposition and its relation to positive definite matrices
USEFUL FOR

Mathematicians, data scientists, and students studying linear algebra who seek a deeper understanding of matrix properties and their implications in various applications.

insixyears
Messages
3
Reaction score
0
nnd = non negative definite matrix
pd = positive definite I'm having a tough time grappling with the concept of pd, psd matrices in general. My understanding basically just boils down to this, basically after multiplying everything out using the matrix formula x'Ax, you will get some sort of polynomial. If you get a polynomial where everything is squared, you're in good shape because it's impossible for the equation to be negative. A question that I have is, what does knowing that a matrix A is psd or pd tell us about the entries of A? Doesn't it only tell us how the entries of A interact (ie, when they are added up, you get something >= 0). If that's true, how would adding the identity matrix give you anymore information about the definiteness of the matrix?Thanks!
 
Physics news on Phys.org
crabchef said:
nnd = non negative definite matrix
pd = positive definite I'm having a tough time grappling with the concept of pd, psd matrices in general. My understanding basically just boils down to this, basically after multiplying everything out using the matrix formula x'Ax, you will get some sort of polynomial. If you get a polynomial where everything is squared, you're in good shape because it's impossible for the equation to be negative. A question that I have is, what does knowing that a matrix A is psd or pd tell us about the entries of A? Doesn't it only tell us how the entries of A interact (ie, when they are added up, you get something >= 0). If that's true, how would adding the identity matrix give you anymore information about the definiteness of the matrix?Thanks!

This is clearly not true. For A to be nonnegative definite, all its entries are nonnegative. What if it has a 0 in any element other than on the main diagonal?
Adding the identity matrix won't change that...
 
crabchef said:
nnd = non negative definite matrix
pd = positive definite I'm having a tough time grappling with the concept of pd, psd matrices in general. My understanding basically just boils down to this, basically after multiplying everything out using the matrix formula x'Ax, you will get some sort of polynomial. If you get a polynomial where everything is squared, you're in good shape because it's impossible for the equation to be negative. A question that I have is, what does knowing that a matrix A is psd or pd tell us about the entries of A? Doesn't it only tell us how the entries of A interact (ie, when they are added up, you get something >= 0). If that's true, how would adding the identity matrix give you anymore information about the definiteness of the matrix?
The definition of nnd is that $x'Ax\geq0$ for every nonzero vector $x$. If that condition holds, then $x'(I+A)x = x'x + x'Ax >0$ (because $x'x>0$), and hence $I+A$ is pd.

Prove It said:
For A to be nonnegative definite, all its entries are nonnegative.
That is not true. The definition of nnd is as given by the OP. It does not imply that all the entries of the matrix are nonnegative.
 
Last edited:
Thanks for the help! As with many problems...AFTER seeing the answer it seems so simple. This gives me one more technique to approach my homework problems, but unfortunately I'm already stuck on the next one (going to keep working on it for a bit longer before I post the question)
 

Similar threads

  • · Replies 9 ·
Replies
9
Views
4K
  • · Replies 15 ·
Replies
15
Views
4K
  • · Replies 4 ·
Replies
4
Views
3K
  • · Replies 5 ·
Replies
5
Views
2K
  • · Replies 26 ·
Replies
26
Views
5K
  • · Replies 19 ·
Replies
19
Views
9K
Replies
2
Views
750
Replies
10
Views
2K
  • · Replies 10 ·
Replies
10
Views
5K
Replies
12
Views
4K