Sei sulla pagina 1di 45

MATH 174: Numerical Analysis I

Jose Marie M. Inaudito


Mathematics Division
Institute of Mathematical Sciences and Physics
University of the Philippines Los Baños

1/9
Announcement (A-4L)

Eleanor B. Gemida
Magna Cum Laude, 2013
2/9
What is Numerical Analysis?

3/9
What is Numerical Analysis?

Ryaben’kii, V.S. and Tsynkov, S.V.


The theory of models and algorithms that are applicable

3/9
What is Numerical Analysis?

Ryaben’kii, V.S. and Tsynkov, S.V.


The theory of models and algorithms that are applicable and that can
be efficiently implemented on computers.

3/9
Symbolic Mathematics vs. Numerical Methods
Symbolic Mathematics

4/9
Symbolic Mathematics vs. Numerical Methods
Symbolic Mathematics
manipulation of variables, functions and expressions

4/9
Symbolic Mathematics vs. Numerical Methods
Symbolic Mathematics
manipulation of variables, functions and expressions
results to exact solutions

4/9
Symbolic Mathematics vs. Numerical Methods
Symbolic Mathematics
manipulation of variables, functions and expressions
results to exact solutions

Numerical Methods
manipulation of numbers

4/9
Symbolic Mathematics vs. Numerical Methods
Symbolic Mathematics
manipulation of variables, functions and expressions
results to exact solutions

Numerical Methods
manipulation of numbers
results to approximations

4/9
Symbolic Mathematics vs. Numerical Methods
Symbolic Mathematics
manipulation of variables, functions and expressions
results to exact solutions

Numerical Methods
manipulation of numbers
results to approximations
uses computers to implement methods

4/9
Ingredients of Numerical Analysis

1 Discretization

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error
4 Methods of Computation

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error
4 Methods of Computation
Accuracy

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error
4 Methods of Computation
Accuracy
Operation Count

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error
4 Methods of Computation
Accuracy
Operation Count
Stability

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error
4 Methods of Computation
Accuracy
Operation Count
Stability
Loss of Significant Digits

5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error
4 Methods of Computation
Accuracy
Operation Count
Stability
Loss of Significant Digits
5 Convergence
5/9
Ingredients of Numerical Analysis

1 Discretization
2 Conditioning
3 Error
Unavoidable Error
Theoretical Error
Round-off Error
4 Methods of Computation
Accuracy
Operation Count
Stability
Loss of Significant Digits
5 Convergence
5/9
Discretization

6/9
Discretization

DISCRETIZATION is the process of converting a problem and/or its key


components to a form a machine can understand.

6/9
Discretization

DISCRETIZATION is the process of converting a problem and/or its key


components to a form a machine can understand.

Types of Discretization

6/9
Discretization

DISCRETIZATION is the process of converting a problem and/or its key


components to a form a machine can understand.

Types of Discretization
Input Discretization

6/9
Discretization

DISCRETIZATION is the process of converting a problem and/or its key


components to a form a machine can understand.

Types of Discretization
Input Discretization involves the conversion of continuous data to a
discrete format.

6/9
Discretization

DISCRETIZATION is the process of converting a problem and/or its key


components to a form a machine can understand.

Types of Discretization
Input Discretization involves the conversion of continuous data to a
discrete format.
Solution Discretization

6/9
Discretization

DISCRETIZATION is the process of converting a problem and/or its key


components to a form a machine can understand.

Types of Discretization
Input Discretization involves the conversion of continuous data to a
discrete format.
Solution Discretization involves the conversion of an infinite or
continuous part of a solution to a finite and discrete set of operations.

6/9
Conditioning

7/9
Conditioning

Consider the problem of finding the roots of x2 + 2αx + 1 via the algorithm
p
A : x = −α + α2 − 1

7/9
Conditioning

Consider the problem of finding the roots of x2 + 2αx + 1 via the algorithm
p
A : x = −α + α2 − 1

For what value of α is A sensitive to?

7/9
Conditioning

Consider the problem of finding the roots of x2 + 2αx + 1 via the algorithm
p
A : x = −α + α2 − 1

For what value of α is A sensitive to?

We have shown that when |α| > 1,

7/9
Conditioning

Consider the problem of finding the roots of x2 + 2αx + 1 via the algorithm
p
A : x = −α + α2 − 1

For what value of α is A sensitive to?

We have shown that when |α| > 1, the problem is well-conditioned,

7/9
Conditioning

Consider the problem of finding the roots of x2 + 2αx + 1 via the algorithm
p
A : x = −α + α2 − 1

For what value of α is A sensitive to?

We have shown that when |α| > 1, the problem is well-conditioned, while
as α gets close to 1, the problem is ill-conditioned.

7/9
Conditioning

Consider the problem of finding the roots of x2 + 2αx + 1 via the algorithm
p
A : x = −α + α2 − 1

For what value of α is A sensitive to?

We have shown that when |α| > 1, the problem is well-conditioned, while
as α gets close to 1, the problem is ill-conditioned.

7/9
Illustration:
The following table illustrates the solutions for equations with varying α ’s
obtained via A

8/9
Illustration:
The following table illustrates the solutions for equations with varying α ’s
obtained via A

Taking enormous leaps

8/9
Illustration:
The following table illustrates the solutions for equations with varying α ’s
obtained via A

Taking enormous leaps


∆x
α x | ∆α | (Relative Change)
1000 −0.0005
999.9 −0.00050005 5.0005 × 10−3
105 −4.99999 × 10−6 5 × 10−9

8/9
Illustration:
The following table illustrates the solutions for equations with varying α ’s
obtained via A

Taking enormous leaps


∆x
α x | ∆α | (Relative Change)
1000 −0.0005
999.9 −0.00050005 5.0005 × 10−3
105 −4.99999 × 10−6 5 × 10−9

Taking micro steps

8/9
Illustration:
The following table illustrates the solutions for equations with varying α ’s
obtained via A

Taking enormous leaps


∆x
α x | ∆α | (Relative Change)
1000 −0.0005
999.9 −0.00050005 5.0005 × 10−3
105 −4.99999 × 10−6 5 × 10−9

Taking micro steps


∆x
α x | ∆α |
1 −1
1.1 −0.641742 3.58258
1.01 −0.868226 86.8226

8/9
Illustration:
The following table illustrates the solutions for equations with varying α ’s
obtained via A

Taking enormous leaps


∆x
α x | ∆α | (Relative Change)
1000 −0.0005
999.9 −0.00050005 5.0005 × 10−3
105 −4.99999 × 10−6 5 × 10−9

Taking micro steps


∆x
α x | ∆α |
1 −1
1.1 −0.641742 3.58258
1.01 −0.868226 86.8226

8/9
2. Conditioning: Well- or Ill-conditioned

CONDITIONING has to do with quantifying the sensitivity of the


solution to perturbations of the input data.

If sensitivity is low, the problem is well-conditioned.


If sensitivity is high, the problem is ill-conditioned.

9/9

Potrebbero piacerti anche