Should we compare floating point numbers for equality against a *relative* error?

by

It all depends on the specific problem domain. Yes, using relative error will be more correct in the general case, but it can be significantly less efficient since it involves an extra floating-point division. If you know the approximate scale of the numbers in your problem, using an absolute error is acceptable.

This page outlines a number of techniques for comparing floats. It also goes over a number of important issues, such as those with subnormals, infinities, and NaNs. It’s a great read, I highly recommend reading it all the way through.

More Related Contents:

  1. How should I do floating point comparison?
  2. Is floating point math broken?
  3. How Does Modulus Divison Work
  4. How do I display the binary representation of a float or double?
  5. Why does division by zero in IEEE754 standard results in Infinite value?
  6. What types of numbers are representable in binary floating-point?
  7. Calculating a 2D Vector’s Cross Product
  8. Why does floating-point arithmetic not give exact results when adding decimal fractions?
  9. String Compare “Logic”
  10. Custom Weighted Scale [closed]
  11. Why are floating point numbers inaccurate?
  12. Determine Whether Two Date Ranges Overlap
  13. Division of integers in Java [duplicate]
  14. What is a magic number, and why is it bad? [closed]
  15. What is the difference between currying and partial application?
  16. What is Type-safe?
  17. Try to describe polymorphism as easy as you can [closed]
  18. Does float have a negative zero? (-0f)
  19. Modulus power of big numbers
  20. Acronyms in CamelCase [closed]
  21. What is the best way to replace or substitute if..else if..else trees in programs?
  22. Shortest distance between two line segments
  23. What is the difference between IF-ELSE and SWITCH?
  24. Python: max/min builtin functions depend on parameter order
  25. Followup: Finding an accurate “distance” between colors
  26. Line of intersection between two planes
  27. JavaScript: Round to a number of decimal places, but strip extra zeros
  28. How to design extensible software (plugin architecture)? [closed]
  29. XOR of three values
  30. Conditional tests in primality by trial division

Leave a Comment