Sample evenly-ish vector of unevenly distributed indices

Say I have a vector of indices

v = [v1 v2 v3 v4 v5] = [1 34 456 1345 4567]

I try to find a way to sample v in an evenly-ish way while keepking these points. Is there a quick way to do that instead of a loop over every range between two points [v_i v_j], I have numerous and long vectors...

[Walter Roberson]

Could you talk more about the sampling process? Is this like you want (say) 5 locations within each segment? How should the endpoints be treated, included in both, left is included but right isn't? Except right is included in the last segment? Details!

[Ouatehaouks]

Perhaps my question is clearer if I say that: I have a vector v, I want to get a new vector vnew in which I have all points of v but also new points so that diff(vnew) does not vary too much.

[Voss]

How about this:

vnew = min(v):max(v);

Now, as long as v contains only integer values, all elements of v are in vnew. And diff(vnew) is all 1s, so the difference is as constant as it can get.

Does that work?

[Ouatehaouks]

Thanks for your answer. The thing is that the indices can be very large, so I want to refine v but not down to delta=1.

Answer 1

[Matt J]

Perhaps as follows,

vnew = union( min(v):delta:max(v) ,v);

[Ouatehaouks]

Thanks- This was also my first try indeed, but with this approach the problem is that diff(vnew) can go down to almost zero if min(v):delta:max(v) generates indices close to those of v. So I think that somehow I need to consider every range between each points of v to get the proper delta and then to generate the points but I need to do that without a loop otherwise it is too long.

[Matt J]

Easy enough to remove those points:

vnew=uniquetol(vnew,1e-8,'DataScale',1)

[Matt J]

Or perhaps, as follows. This ensures that v is preserved within vnew exactly

 vnew=min(v):delta:max(v);
 
 tf=ismembertol(vnew,v,tol,'DataScale',1);
 
 vnew(tf)=[];
 
 vnew=union(vnew,v);

[Ouatehaouks]

I did not know about ismembertol, this is it ! That way I can indeed make sure I get rid of points that are too close from those of v. Thank you for your help !

[Ouatehaouks]

Sorry I don't know how to accept the correct answer which is in your last comment.

[Matt J]

No problem. Our whole chain of comments is considered part of the "Answer".