Scribd
Upload
76 views

Quadratic Forms and Local Extrema: JLB MAT3216: Handout 2

This document discusses quadratic forms and using the second derivative test to classify local extrema of functions. It defines quadratic forms as functions of the form q(x) = xT Hx, where H is a symmetric matrix. Quadratic forms can be positive definite, negative definite, or indefinite depending on the signs of q(x) for all x. The second derivative test classifies critical points of functions as local minima, maxima or saddle points based on the definiteness of the Hessian matrix. For quadratic forms defined by 2x2 matrices, the definiteness depends on the signs of the diagonal entries and the determinant of the matrix.

Uploaded by

SophaVisa Khun
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
76 views

Quadratic Forms and Local Extrema: JLB MAT3216: Handout 2

This document discusses quadratic forms and using the second derivative test to classify local extrema of functions. It defines quadratic forms as functions of the form q(x) = xT Hx, where H is a symmetric matrix. Quadratic forms can be positive definite, negative definite, or indefinite depending on the signs of q(x) for all x. The second derivative test classifies critical points of functions as local minima, maxima or saddle points based on the definiteness of the Hessian matrix. For quadratic forms defined by 2x2 matrices, the definiteness depends on the signs of the diagonal entries and the determinant of the matrix.

Uploaded by

SophaVisa Khun
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Quadratic Forms and Local Extrema

Let H be a symmetric matrix. Then the function q : Rn → R defined by

q(x) = xT Hx
1 −1

is called a quadratic form. For example if H = −1 1 then
  
 1 −1 x
q(x, y) = x y = x2 − 2xy + y 2 .
−1 1 y

Another example is the third term in the Taylor expansion of a scalar-valued


function on Rn :
 
D1 D1 f (a) · · · Dn D1 f (a)
1 1  .. ... ..
q(h) = hT Hf (a)h = hT  h

2 2 . .
D1 Dn f (a) · · · Dn Dn f (a)
n
1X
= hi hj Dj Di f (a).
2 i,j=1

Definition: Let q : Rn → R be a quadratic form. Then q is said to be

(a) positive definite if q(x) > 0 for all x ∈ Rn ;

(b) negative definite if q(x) < 0 for all x ∈ Rn ;

(c) indefinite if q(x) > 0 for some x ∈ Rn and q(x) < 0 for some x ∈ Rn .

The symmetric matrix H is also sometimes said to be positive definite if its corre-
sponding quadratic form is positive definite etc.

We can now generalize the second derivative test for classifying local extrema.

Theorem (Second Derivative Test): Let f : A ⊆ Rn → R be of class


3
C (A), where A is open. Let a be a critical point of f . Then

(i) if the Hessian of f is positive definite, then a is a local minimum of f ;

(ii) if the Hessian of f is negative definite, then a is a local maximum of f ;

(iii) if the Hessian of f is indefinite, then a is a saddle point.

JLB 1 MAT3216: Handout 2


We note that a matrix is positive definite if all its eigenvalues are positive, neg-
ative definite if all its eigenvalues are negative, and indefinite if it has both positive
and negative eigenvalues. (Recall that all the eigenvalues of a symmetric matrix
are real, and it is always possible to find an orthonormal basis of eigenvectors.)

The following criterion helps us classify quadratic forms that arise from 2 × 2
matrices.
 
a b
Proposition: Let H be the 2 × 2 symmetric matrix . Then the
b d
quadratic form q(x) = xT Hx is

(i) positive definite if a > 0 and det(H) > 0;

(ii) negative definite if a < 0 and det(H) > 0;

(iii) indefinite if det(H) < 0.

Combining this with the above theorem, we get the following result.

Theorem : Let f : A ⊆ R2 → R be of class C 3 (A), where A is open. Then a


is a (strict) local maximum of f if

(i) D1 f (a) = D2 f (a) = 0 (i.e. a is a critical point);

(ii) D1 D1 f (a) < 0;



D1 D1 f (a) D2 D1 f (a)
(iii) D = > 0.
D1 D2 f (a) D2 D2 f (a)

If D1 D1 f (a) > 0, then a is a strict local minimum, while if D < 0, then a is a


saddle point.

Note that if D = 0, the test fails.

JLB 2 MAT3216: Handout 2

You might also like