0% found this document useful (0 votes)
48 views23 pages

Lecture 6 Optimum Design

The document discusses Taylor series expansions for functions of single and multiple variables. It explains that Taylor series can be used to approximate functions around a point and derive necessary conditions for identifying an optimum, such as a minimum or maximum. The first-order necessary condition is that the first derivative must be equal to zero at the optimal point. The second-order sufficient condition is that the second derivative must be positive at a minimum and negative at a maximum. Higher order derivatives may also need to be considered if the second derivative is equal to zero.

Uploaded by

Mohmmad Break
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
0% found this document useful (0 votes)
48 views23 pages

Lecture 6 Optimum Design

The document discusses Taylor series expansions for functions of single and multiple variables. It explains that Taylor series can be used to approximate functions around a point and derive necessary conditions for identifying an optimum, such as a minimum or maximum. The first-order necessary condition is that the first derivative must be equal to zero at the optimal point. The second-order sufficient condition is that the second derivative must be positive at a minimum and negative at a maximum. Higher order derivatives may also need to be considered if the second derivative is equal to zero.

Uploaded by

Mohmmad Break
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/ 23

OPTIMUM DESIGN

Dr. / Ahmed Nagib Elmekawy

Lecture 6

Fall 2017

1
Taylor Series Expansion
Assume f(x) is:
1. Continuous function of a single variable x
2. Differentiable n times
3. x ∈ S, where S is non-empty, closed, and bounded
4. therefore x* is a possible optima

df ( x*) 1 d 2 f ( x*) 1 d 3 f ( x*)


f ( x )  f ( x*)  ( x  x*)  2
( x  x*) 2
 3
( x  x*) 3
 h.o.t.
dx 2 dx 3! dx

2
Taylor Series Approximations
let x  x*  d , a small change or vicinit y of x * ...then
f ( x )  f ( x * d )
df ( x*) 1 d 2 f ( x*) 2 1 d 3 f ( x*) 3
f ( x )  f ( x*)  d 2
d  3
d  h.o.t.
dx 2 dx 3! dx
df ( x*)
f ( x )  f ( x*)  d First Order Approximation
dx
df ( x*) 1 d 2 f ( x*) 2
f ( x )  f ( x*)  d 2
d Second Order Approximation
dx 2 dx

3
Taylor Series – Mult. Variables
Estimate f(x) using Taylor Series
1
f (x)  f (x*)  f (x*)( x  x*)  (x  x*)T H(x*) (x  x*)  R
T
2

Let d  (x  x*) , theref ore x  x *  d


1 T
f ( x )  f (x * d)  f (x*)  f (x*)d  d H d  R
T
2

4
Change in function

Let’s define a change in the function with the symbol ∆

f  f (x )  f (x*)
1 T
f  f d  d H d  R
T

2
∆-Delta ∇-Del

Let’s define a first order change in the function


δf  f T δx  f • δx

5
Taylor Exp of MV

6
7
8
Taylor Series Expansion
Assume f(x) is:
1. Continuous function of a single variable x
2. Differentiable n times
3. x ∈ S, where S is non-empty, closed, and bounded
4. therefore x* is a possible optima

df ( x*) 1 d 2 f ( x*)
f ( x )  f ( x*)  ( x  x*)  2
( x  x*)2
 ...
dx 2 dx
1 d 3 f ( x*) 1 d 4 f ( x*)
3
( x  x*)3
 4
( x  x*)4
 h.o.t.
3! dx 4! dx

9
Taylor Series Approximations
let x  x*  d , a small change or vicinit y of x * ...then
f ( x )  f ( x * d )
df ( x*) 1 d 2 f ( x*) 2 1 d 3 f ( x*) 3
f ( x )  f ( x*)  d 2
d  3
d  h.o.t.
dx 2 dx 3! dx

df ( x*)
f ( x )  f ( x*)  d First Order Approximation
dx
df ( x*) 1 d 2 f ( x*) 2
f ( x )  f ( x*)  d 2
d Second Order Approximation
dx 2 dx

10
Single variable minimization
Given that x* is the minimum of f(x), then any
movement away from x* is “uphill”, therefore to
guarantee that a move goes uphill
f ( x )  0
1
f ( x )  f ( x )  f ( x*)  f ( x*)d  f ( x*)d 2  ...
2
1 1
f ( x*)d 3  f iv ( x*)d 4  h.o.t.
3! 4!

First-order necessary condition


f ( x*)  0

11
Stationary point=max,min,neither
Any points satisfying f ( x*)  0

Are called “stationary” points.

Those points are a:


1. Min pt, or
2. Max pt, or
3. Neither (i.e. an inflection pt)

We need another test!

12
Second-order sufficient condition

Look at second order term


1
f ( x )  f ( x )  f ( x*)  f ( x*)d  f ( x*)d 2  ...
2
1 1
f ( x*)d 3  f iv ( x*)d 4  h.o.t.
3! 4!

Second-order sufficient condition for a minimum


f ( x*)  0

13
Second-order condition?

What if f ( x*)  0


Then f(x*) is not a minimum of f(x*).

It is a maximum of f(x*).

14
Single variable optimization
First-order necessary condition

f ( x*)  0

f ( x*)  0 f ( x*)  0


Second-order sufficient Second-order sufficient
condition for a minimum condition for a maximum
? ?
15
Higher—order tests?
What if f ( x*)  0

Then the second order test fails. We need


higher order derivatives…

Min f even ( x*)  0

Max f even ( x*)  0

16
Second-order necessary conditions

Note the “= 0” possibility


f ( x*)  0 f ( x*)  0

A pt not satisfying this A pt not satisfying this


test is not a min! test is not a max!

17
Multiple variable optimization
If x* is the minimum of f(x), then any movement
away from x* is “uphill”.

f  f (x )  f (x*)  0
1 T
f  f (x * d)  f (x*)  f (x*)d  d H d  R
T

2
1 T
f  f (x*)d  d H d (second order)
T

2
dot product triple product
scalar scalar
f  scalar
How can we guarantee that for a move in any d, we
make away from x*, we go “uphill”?
18
First-order necessary Condition
For x* to be a local minimum: f  f (x )  f (x*)  0
1 T
f  f (x*)d  d H d
T

2
 f 
 x  0
1rst order term 0
 1
 f   
  0
f (x*)  0
T
f (x*)  x2  0
  0
   
 f   
   

 n  x*
x 0

19
Second-order sufficient condition

For x* to be local minimum:

dT H d  0

That is H(x*) must be positive definite

Remember that dT H d

has “quadratic form”

20
Quadratic form of a matrix
1 T
Given F (x )  x A x where A is symmetric, for example :
2
1 0 0  x1 
xT  x1 x2 x3  A  0 1 0 x   x2 
   x3 
0 0 1
1 0 0  x   x 
  1
x A x  x • (Ax ) Ax  0 1 0  x2    x2 
T T 1

 x  x 
0 0 1  3   3 
 x1 
x A x  x • ( Ax )  x1 x3   x2   x12  x22  x32
T T T
x2
 x3 
F (x )  x1  x22  x32 
1 2
2 21
For example, the following matrices are also associated with the same quadratic
form:

22
Dividing the coefficients equally between the off-diagonal terms, we obtain the symmetric
matrix associated with the quadratic form in E q. (a) as

23

You might also like