For loop versus Matrix notation

Can this be simplified to use a matrix expression instead of a for loop?:

X=[1 5.4; 1 6.3; 2 4.8; 3 7.1];
dates=X(:,1);
amounts=X(:,2);
uniquedates=unique(dates);
totals=zeros(size(uniquedates));
for d = 1:size(uniquedates,1)
    totals(d,1)=sum(amounts(dates==uniquedates(d,1)));
end
Y = [uniquedates totals];

[Matt Kindig]

Probably, but likely not in an easy-to-understand way. The fact that totals differs in size from dates complicates things.

Is there a reason you don't want to just use the for-loop?

[mutt]

I want to find out if there is a matrix approach which can outperform the for loop when the inputs are non-trivial

Answer 1

[Oleg Komarov]

You can use accumarray():

 Y = [uniquedates accumarray(X(:,1),X(:,2))];

[mutt]

What if the values in X(:,1) are times and therefore non-integer?

[Oleg Komarov]

Map the dates to a set X in the positive N, i.e. use unique on the dates and then X(:,1) is the ia index from the call to unique.

[mutt]

This approach has delivered a 99% time saving versus the for loop.