Matrix/Vector Differentiation: Proving the Derivative of x'Mx

  • Thread starter Thread starter wu_weidong
  • Start date Start date
  • Tags Tags
    Differentiation
Click For Summary
SUMMARY

The discussion focuses on proving the derivative of the function g(x) = x'Mx, where M is an n-by-n real constant matrix. The correct derivative is established as (M + M')x, countering the initial incorrect approach using the product rule. The user Rayne seeks clarification on the differentiation process, specifically regarding the dimensions of the terms involved. The solution involves recognizing the summation indices and applying matrix differentiation rules accurately.

PREREQUISITES
  • Understanding of matrix calculus, specifically differentiation of scalar functions with respect to vector variables.
  • Familiarity with matrix transposition and properties of symmetric matrices.
  • Knowledge of the product rule in the context of matrix operations.
  • Ability to manipulate summation indices in tensor notation.
NEXT STEPS
  • Study matrix differentiation techniques, focusing on scalar functions of vector variables.
  • Learn about the properties of symmetric matrices and their implications in differentiation.
  • Explore the application of the product rule in matrix calculus.
  • Investigate tensor notation and its use in simplifying matrix expressions.
USEFUL FOR

Students and professionals in mathematics, engineering, and physics who are involved in advanced calculus, particularly those working with matrix differentiation and optimization problems.

wu_weidong
Messages
27
Reaction score
0

Homework Statement



Hi all,
I need help proving the result:

Let g(x) = x'Mx, where M is a n-by-n real constant matrix and x' denotes the transpose of vector x. Then the derivative of g(x) = (M + M')x.

The Attempt at a Solution



I was thinking of using product rule on x'(Mx) to get Mx + x'M, but apparently this is incorrect as the dimensions of Mx and x'M don't even match and so cannot be grouped together to get (M + M')x.

Please help.

Thank you.

Regards,
Rayne
 
Physics news on Phys.org
Think indices. x'Mx=x_i*M_ik*x_k (summed over i and k). What's the derivative of that wrt, say, x_n?
 

Similar threads

  • · Replies 8 ·
Replies
8
Views
1K
Replies
5
Views
2K
  • · Replies 22 ·
Replies
22
Views
3K
  • · Replies 8 ·
Replies
8
Views
2K
Replies
10
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 10 ·
Replies
10
Views
2K
  • · Replies 39 ·
2
Replies
39
Views
6K
  • · Replies 6 ·
Replies
6
Views
2K
  • · Replies 4 ·
Replies
4
Views
4K