1 / 33

Error Analysis Part 1 The Basics

Error Analysis Part 1 The Basics. Key Concepts. Analytical vs. numerical Methods Representation of floating-point numbers Concept of significant digits Distinguishing different kinds of errors Round-off / chopping / truncation errors True/approximate absolute and relative errors

dakota
Télécharger la présentation

Error Analysis Part 1 The Basics

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Error Analysis Part 1 The Basics

  2. Key Concepts • Analytical vs. numerical Methods • Representation of floating-point numbers • Concept of significant digits • Distinguishing different kinds of errors • Round-off / chopping / truncation errors • True/approximate absolute and relative errors • Acceptable errors

  3. Analytical vs. Numerical Methods • Find the intersection of y1 = 2x + 3 y2 = x + 2 • Find the intersection of y1 = x y2 = cos(x)

  4. Analytical vs. Numerical Methods • Analytical Methods • Accurate solution • Difficult and not always possible • Numerical Methods • Approximation of true solution • What method to use? • How good is our approximation? (Error Analysis) • How efficient is our method? (Algorithm design, Convergence rate) • Does our methods always work? (Convergence)

  5. Number Representation • Do machines represent integers and floating-point numbers using the same representation? • How does computer represent integers? • How does computer represent floating-point numbers?

  6. Representation of Integers 13 as 8-bit unsigned integers (no negative #) 1310 = 000011012 = 0 x 27 + 1 x 26 + 0 x 25 + 0 x 24 + 1 x 23 + 1 x 22 + 0 x 21 + 1 x 20 = 8 + 4 + 0 + 1

  7. Exercise What is the equivalent decimal number represented by the following binary number? 1101012 = ?

  8. Representation of Floating-point Numbers 156.78 = 0.15678 x 103 in an "imaginary" base-10 floating-point system + 3 15678

  9. Normalized Representation(and notations used in this course) • σ is the sign • β is the base, e is the exponent • binary : β =2 • decimal : β =10 • 1/β≤ m < 1 (i.e., a1 ≠ 0) • binary: 0.5 ≤ m < 1 • decimal: 0.1 ≤ m < 1

  10. Representation of Floating-point Numbers

  11. Exercise • What is the normalized floating-point representation of 12.7510 (for β = 2)? • What is the normalized floating-point representation of 2.210 (for β = 2)? • What is the equivalent decimal value of (0.110111)2 x 23 ?

  12. There are discrete points on the number lines that can be represented by our computer. How about the space between ?

  13. Implication of floating-point representations • Only limited range of quantities may be represented • Number too larger  overflow • Number too small (too close to 0) underflow • Only a finite number of quantities may be represented • round-off or chopping errors

  14. Exercise • Consider the following floating-point representation • The mantissa has only 3 bits • Exponent, e, ranges from -4 to 4 • Can you give an integer that cannot be represented by this representation? • Can you give an integer between 0 and 14 that cannot be represented by this representation?

  15. IEEE 754 Floating-point Representation Larger exponent  Wider range of numbers Longer mantissa  Higher precision

  16. Note on IEEE 754 Representation • Exponents of all 0's and 1's are reserved for special numbers. • Zero is a special value denoted with an exponent field of zero and a mantissa field of zero, and we could have +0 and -0. • +∞ an -∞ are denoted with an exponent of all 1's and a mantissa field of all 0's. • NaN (Not-a-number) is denoted with an exponent of all 1's and a non-zero mantissa field.

  17. Errors and Significant Digits • I paid $10 for 7 oranges. What is unit price of each orange? • $1.428571429 (that is the exact output from my computer !!) • Is there any difference between $1.427571429 and $1.4? • Is there any difference between $1.4 and $1.40?

  18. Significant figures, or digits • The significant digits of a number are those that can be used with confidence. • They correspond to the number of certain digits plus one estimated digits. • x = 3.5 (2 significant digits)  3.45≤ x < 3.55 • x = 0.51234 (5 significant digits)  0.512335≤ x < 0.512345

  19. Excercise • Suppose x = 3.141592658979323 • Show the value of x up to 4 significant digits. • Show the value of x up to 10 significant digits. • Calculate 22/7 up to 5 significant digits.

  20. Concepts of Significant Digits Supposex = 0.739085 is the true solution Which of the following calculated values is/are accurate to 3 significant digits with respect to x? a = 0.739505 b = 0.739626 c = 0.739379 d = 0.738999

  21. Concepts of Significant Digits xA (approximate value) has m significant digits (with respect to xT, the true value) if the absolute error | xT - xA | has magnitude less than or equal to 5 in the (m + 1)st digit of xT counting to the right from the first non-zero digit in xT. e.g. 1:  3 significant digits

  22. e.g. 2:  4 significant digits e.g. 3:  2 significant digits

  23. Excercise Supposex = 0.739085 is the true solution Which of the following calculated values is/are accurate to 3 significant digits with respect to x? a = 0.739505 (| x - a | = 0.000420) b = 0.739626 (| x - b | = 0.000541) c = 0.739379 (| x - c | = 0.000294) d = 0.738999 (| x - d | = 0.000086)

  24. Scientific Notation How do we express the number 45,300 meaningfully? • 4.53 x 104 to denote the number is known to 3 significant figures. • 4.530 x 104to denote the number is known to 4 significant figures. • 4.5300 x 104 to denote the number is known to 5 significant figures.

  25. Implications As numerical methods yield approximate results, we must therefore develop criteria to specify how confidence we are in our approximate result. Usually, in terms of 1) Significant digits, or 2) Absolute/relative error bounds

  26. Error Definition (True Error) xT – true value xA – approximate value True Error in xA (exact value of the error) = True Relative Error in xA= True Percentage Relative Error in xA=

  27. Error Definition • e.g., • Error = • Relative error =

  28. Error Definition (Approximate Error) • If we do not know the true value xT, we can replace it by an estimation of the true value. • The result is, we have the approximate error, and the approximate relative error instead.

  29. Error Definition (Approximate Error) xA(i) – approximate value in the ith iteration of an iterative approach Approximate Error in xA = Approximate Relative Error in xA = Approximate Percentage Relative Error in xA

  30. Example: Maclaurin Series When x = 0.5

  31. How many terms should we use? • Computation stops when |εa| < εs • εs = pre-determined acceptable percentage relative error • If we want the result to be correct to at least n significant digits, it is suggested that we set εs = (0.5 x 102-n)%

  32. Summary • Floating-point number representation and its implication • Round-off and chopping errors • Significant digits • The definitions of • True errors, true relative errors, true percentage errors, • Approximate errors, approximate relative errors, approximate percentage relative errors

  33. Next Errors do not occur only in the space between the discrete values (rounding or chopping error) Errors also appear in many stages. Propagation of round-off errors Truncation errors

More Related