Help with full rank factorization

  • Context: Graduate 
  • Thread starter Thread starter learningstill
  • Start date Start date
  • Tags Tags
    Factorization rank
Click For Summary
SUMMARY

The discussion focuses on proving the existence of a full rank factorization for an arbitrary m x n matrix A, where rank(A) = r. It establishes that there exist matrices B ∈ R^{m x r} and C ∈ R^{r x n} such that A = BC, and confirms that rank(A) = rank(B) = r. The user seeks guidance on proving these properties, particularly leveraging the relationship rank(AB) ≤ rank(A) and rank(AB) ≤ rank(B).

PREREQUISITES
  • Understanding of matrix rank and its properties
  • Familiarity with linear transformations and their representations
  • Knowledge of kernel and image of a matrix
  • Basic concepts of vector spaces and direct sums
NEXT STEPS
  • Study the proof of the rank-nullity theorem
  • Explore the implications of the singular value decomposition (SVD)
  • Learn about the properties of linear transformations in vector spaces
  • Investigate applications of full rank factorization in numerical methods
USEFUL FOR

Mathematicians, data scientists, and students studying linear algebra, particularly those interested in matrix factorization and its applications in computational mathematics.

learningstill
Messages
3
Reaction score
0
I've been tasked with proving the existence of a full rank factorization for an arbitrary m x n matrix, namely:

Let \textit{A} \in \textbf{R}^{m x n} with \textit{rank(A) = r} then there exist matrices \textit{B} \in \textbf{R}^{m x r} and \textit{C} \in \textbf{R}^{r x n} such that \textit{A = BC}. Furthermore \textit{rank(A) = rank(B) = r}.

I think I can prove the second property if I assume the first using \it{rank(AB)} \leq \it{rank(A)} and \it{rank(AB)} \leq \it{rank(B)}.

I'd appreciate a push in the right direction. Thanks.
 
Last edited by a moderator:
Physics news on Phys.org
We have the following situation:
$$
V_n \stackrel{C}{\twoheadrightarrow} V_n/\operatorname{ker}A \cong V_r \cong \operatorname{im}A \stackrel{B}{\rightarrowtail} V_m
$$
which together ##BC## transform as ##A##. We have split the domain of ##A\, : \,V_n \longrightarrow V_m ## into ##V_n \cong \operatorname{ker}A \oplus V_n/\operatorname{ker}A## and the codomain in ##\operatorname{im}A \oplus V_{m-r}\,.##
 

Similar threads

  • · Replies 24 ·
Replies
24
Views
1K
  • · Replies 0 ·
Replies
0
Views
2K
  • · Replies 1 ·
Replies
1
Views
3K
  • · Replies 1 ·
Replies
1
Views
51K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 6 ·
Replies
6
Views
7K
  • · Replies 8 ·
Replies
8
Views
5K
  • · Replies 1 ·
Replies
1
Views
3K
Replies
1
Views
9K