Can this for-loop code get faster in some way?

1 次查看（过去 30 天）
Martin
Martin 2019-10-5
评论： darova 2019-10-6
I got a big result table named resTbl.
There I need for each row to grab a timestamp (posix time) and construct a period interval vector (I use a constant, ts_length, to construct this).
Then I need to find all those in period_interval that are represented in the variable data (column 2), and take the sum of those in data (column 7).
The below code works as I want it to:
for i = 1 : size(resTbl ,1)
period_interval = resTbl(i,2) : 60000 : resTbl(i,2) + ts_length;
[hd, he] = ismember(period_interval,data(:,2));
resTbl(i,10) = sum(data(he(hd),7));
end
The problem is that it is slow since resTbl has many rows. Does anyone have a suggestion how to make it faster?

请先登录，再回答此问题。

回答（1 个）

darova
darova 2019-10-5
编辑：darova 2019-10-5
Maybe ismember function can be replaced:
dt = 60000;
period_interval = 0 : dt : ts_length;
n = length(period_interval);
for i = 1 : size(resTbl ,1)
cond = ~mod( data(:,2)-resTbl(i,2),dt ); % multiple by 60 000
mult = (data(:,2)-resTbl(i,2))/dt;
ind = (0 <= mult & mult <= n) & cond; % (0 <= multiplier <= n) and multiple by 60 000
% [hd, he] = ismember(period_interval,data(:,2));
resTbl(i,10) = sum(data(ind,7));
end
  4 个评论
显示 2更早的评论
Martin
Martin 2019-10-6
Pretty brilliant solution I really have to admit!
It is faster but unfortunately it is however also pretty time consuming and VERY heavy on the memory due to the matrices!
With my current memory, 32 GB, I can not run the full resTbl set (3400 rows) on the dataset which is like 2 million rows and 7 columns.
I tried with smaller resTbl and data set, and for some reason I had to alter n to this: n = length(period_interval)-1; to match my own results.
I am not even sure if there exist a better solution to this problem than yours. I gladly hear from you again, but otherwise I say thank you very much!
darova
darova 2019-10-6
2 million rows and 7 columns
Maybe time is a price in this case

请先登录，再进行评论。

类别

在 Help Center 和 File Exchange 中查找有关 Logical 的更多信息

标签

产品

Community Treasure Hunt

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

Start Hunting!

Translated by