	Department of Engineering

University of Cambridge > Engineering Department > computing help > programs > matlab

Matlab - NaN and Inf

Introduction

If in matlab you type 1/0 you'll get

  Warning: Divide by zero.
  (Type "warning off MATLAB:divideByZero" to suppress this warning.)

  ans =
   Inf

which is fair enough - it's potentially useful to for 1/0 to result in infinity. If you type 0/0 you'll get

  Warning: Divide by zero.
  (Type "warning off MATLAB:divideByZero" to suppress this warning.)

  ans =
   NaN

NaN means "Not a Number" - i.e. the result is undefined. Both these special values behave in predictable ways. Any mathematical operation involving a NaN results in a NaN. And not surprisingly, most mathematical operation involving Inf result in Inf - Inf/0 and Inf^2 produce Inf. But Inf/Inf is NaN. So is Inf-Inf. isnan and isinf can be use to detect these special values. E.g.

  if (any(isnan(m)))
    disp('NaN values in m')
  end
  if (any(isinf(m)))
    disp('Inf values in m')
  end

Note that

NaN==NaN is false - you need to use isnan to check if a value is NaN
NaN~=NaN is true - NaN isn't a value so it can't be equal to anything
Inf==Inf is true
Inf~=Inf is false

To remove NaN and Inf elements from a vector m you can use

 
  m=m(finite(m))

To replace the NaNs by 0s you can use

 
  m(isnan(m))=0

Graphics

NaNs are ignored when plotting (so are Infs). This is useful if you want to make holes in surfaces

  z = peaks(20); 
  z(4:8,4:8) = NaN*z(4:8,4:8);
  surf(z)

| computing help | Matlab |