# Matlab is making small errors during computations

3 views (last 30 days)
Seeker on 20 Jun 2013
I am sure this question was already asked somewhere, but I was unable to find the answer.
In my code, Matlab is making very small errors during the basic computations.
For example I know that the result of the certain calculation should be equal to 0 and matlab is giving me some extremely small number (2.1E-19) instead.
This is the result of A-B. When I check A and B they appear to be equal but I am sure they differ in 19-th decimal place.
I am sure this is due to the way matlab stores numbers.
What is the best way to deal with it?
Many thanks!

random09983492 on 20 Jun 2013
If you need to compare two numbers, you can use a tolerance. For example, instead of using a == b, use something like abs(a-b) < 1e-10.

Anand on 20 Jun 2013
You haven't given much information, but that's just floating point precision for you. There's tonnes of articles describing this, but the gist of it is that computers use a finite number of bits (and therefore digits) to represent numbers. So when you perform arithmetic operations on these numbers, you will not get 'exact' answers. For most numerical applications this is sufficient.
If this is not sufficient for you, consider using symbolic math.

Seeker on 20 Jun 2013
I am writing a code, where I am doing many computations. I was thing about "forcing" certain values to 0.
Should do the trick..