Uneven data multiplication in Matlab

I have following

Data.numberOfDishes 
Data.numOfLocations

As follows

numberOfDishes  numberOfLocations
3               [1, 2, 4]
4               [9, 8, 5] 
6               [11, 1, 9]   
7               [2, 3, 4]
9               [7, 7, 7]

I want to have resulting column 'total' by doing numberOfDishes/each element of numberOfLocations

For example, so that 'total' for first row would be [3/1, 3/2, 3/4]

i know '.' would not work. mrdivide does not seem an option. Can i do this without having to create numberOfDishes of the same size as numOfLocations

Answer 1

[Star Strider]

This works:

NumberOfDishes = [3 4 6 7 9];
NumberOfLocations = [1, 2, 4;  9, 8, 5;  11, 1, 9;  2, 3, 4;  7, 7, 7];
for k1 = 1:length(NumberOfDishes)
    Total(k1,:) = NumberOfDishes(k1)./NumberOfLocations(k1,:);
end

% To express them as fractions: 
Total = rats(Total)

[Neesha]

Thanks. i would love to avoid for loop, whereever i can but looks like that is the only option here

[Star Strider]

My pleasure!

I certainly agree that it’s nice to vectorise wherever possible. However loops really aren’t evil at all, and are sometimes the most efficient option.

[Image Analyst]

Let's compare the "vectorized" approach with the brute force "for loop" approach:

NumberOfDishes = [3; 4; 6; 7; 9]
NumberOfLocations = [1, 2, 4;  9, 8, 5;  11, 1, 9;  2, 3, 4;  7, 7, 7]
tic
for k1 = 1:length(NumberOfDishes)
    Total(k1,:) = NumberOfDishes(k1)./NumberOfLocations(k1,:);
end
toc
% To express them as fractions: 
Total
Total = rats(Total)

% Use a "vectorized" approach.
tic
t = repmat(NumberOfDishes(:), [1, size(NumberOfLocations, 2)]) ./ NumberOfLocations
toc

Results in command window:

Elapsed time is 0.000032 seconds.
Elapsed time is 0.000402 seconds.

Hmmmm.... Looks like the for loop is 12 times faster.

[Star Strider]

Interesting!

I think the delay with the vectorised approach is the overhead in repmat.

Answer 2

[Andrei Bobrov]

out = bsxfun(@rdivide,NumberOfDishes,NumberOfLocations);