what is the difference between using the zeros array or nan array ?!

Hello! I would like to know what is the difference between initializing an array using the zeros array and nan array in matlab? is there any advantages of using one of them A or B? Thank you!

A= zeros (100,100);
B = nan (100,100);

Answer 1

[John D'Errico]

No difference in theory. I frequently use NaNs because that way, if I am doing something, and then want to verify I have filled the entire array, NaNs would be easier to spot.

zeros seems to be faster, and consistently so. This is certainly due to the way they are implemented internally.

timeit(@() zeros(1000))
ans =
    0.0030123

timeit(@() nan(1000))
ans =
    0.0044756

So, is the time nearly inconsequential? Yes.

Is the difference significant? Yes. NaN takes roughly 50% more time than zeros.

Do I care, unless I have a vast number of calls to either of these operations? NO. And since I try NEVER to write code that has a vast number of calls to any such operator, there is little to worry about for me.

[Steven Lord]

From a non-technical standpoint, if zero is a meaningful value for the quantity you want to store in the array, preallocating with zeros means you can't distinguish between an element that has been filled with 0 and an element that hasn't been filled. For example, consider an array that will store temperatures. In that case, you may want to preallocate using NaN or Inf instead.

[John D'Errico]

What I was trying to say, but Steve said it better. NaN (or inf) is a good way to flag if you did something wrong, especially if zero is possible.