How do I round to the nearest arbitrary, enumerated value in MATLAB?

If I have values:
v = [0.1 2.2 1.6 7.3 10];
and I want all my values to be rounded to the nearest value in this list:
roundTargets = [0 2.7 8 11.1];
then I would want to make a function that would convert each value in v to the nearest value in roundTargets.
So this function would return me:
vRounded = [0 2.7 2.7 8 11.1];
Any clever ways to do this?
[Yes, I know how to do this. When I get questions from users that make good posts here, I like to put them out for the world to find.]

3 Comments

What I love about this is there are three perfectly good answers, all doing this in a slightly different way. Each of them has the possibility to be the best solution under different constraints (memory, time, ease of development, etc...)
Would you mind providing those 3 solutions? :)
The three solutions are shown below -- interp1, histc, bsxfun
Note: Doug Hull moved on to a different career so he will not be back to respond.

Sign in to comment.

 Accepted Answer

One approach:
roundTargets = [0 2.7 8 11.1];
v = [0.1 2.2 1.6 7.3 10];
vRounded = interp1(roundTargets,roundTargets,v,'nearest')

3 Comments

Exactly! Now, can you think of key words to put into this solution so other people can find it? That is the real art here, no?
I will add some tags, as I think that helps the search most.
To account for values outside the span of roundTargets, you would need
vRounded = interp1(roundTargets,roundTargets,v,'nearest','extrap')

Sign in to comment.

More Answers (2)

I believe HISTC is the fastest and most memory efficient way to solve this problem for large cases.
v = 10*rand(1,1000000);
roundTargets = [0 sort(10*rand(1,10000)) 10];
%%%histc
tic
[~,Index1] = histc(v,[-Inf interp1(1:numel(roundTargets),roundTargets,0.5 + (1:numel(roundTargets)-1)) Inf]);
vRounded1 = roundTargets(Index1);
timehistc = toc
%%%interp1
tic
vRounded2 = interp1(roundTargets,roundTargets,v,'nearest');
timeinterp1 = toc
isequal(vRounded1,vRounded2)
%%%bsxfun will run out of memory
tic
[~,idx] = min(bsxfun(@(x,y)abs(x-y),v,roundTargets.')); %index of closest
vRounded = roundTargets(idx); %extract values
timebsxfun = toc
[~,idx] = min(bsxfun(@(x,y)abs(x-y),v,roundTargets.')); %index of closest
vRounded = roundTargets(idx); %extract values
Keywords?
minimum, absolute, difference, element by element, comparison, index

1 Comment

You are too fast! I was about to post this with the warning that it could take a ton of memory for large problems because of the intermediate array.

Sign in to comment.

Products

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!