Question

I have a vector which is normalize(i.e. the sum of its elements is 1) and they are ordered descending(i.e. [0.5 0.4 0.09 0.01]).

Is there a way of obtaining the indices of the first elements who’s sum is just bellow a certain threshold?

For example, in my case, for a threshold of 0.6 I would obtain 1, the index(w.r.t. to Matlab way of indexing vector elements) of 0.5. For a threshold of 0.91 I would obtain [1 2], the indices of [0.5 0.4], and so on.

I know that I can do this with a loop through the vector but given the fact that I am using Matlab, I thought asking whether I can do this with just a single command line or max 2, thus increasing the computational speed of my code.

Answer 1

The cumsum function computes the cumulative sum of a vector’s elements. doing this, and then finding the index for which the threshold turns smaller than the cumulative sum will give you your answer:

V = [0.5 0.4 0.09 0.01]

t = 0.6;

index = find(t < cumsum(V), 1)-1

Note that if index = 0, the condition does not hold even for the first element.

If you insist on getting a vector of elements that qualify, you can simply define a range with this:

indices = 1 : (find(t<cumsum(V),1)-1)

In this case, the indices vector is empty in case the first element doesn’t qualify.