# How can i do it fast way ?

25 views (last 30 days)
mankang on 18 Feb 2021
Edited: mankang on 19 Feb 2021
example = mat2cell(rand(100000,4),ones(1,100)*1000,4);
result = cell( size(example, 1), 1);
var1_list = [0.5, 0.7];
var2_list = [0.3, 0.5, 0.7];
final = cell( length(var1_list)*length(var2_list), 1);
cnt = 1;
for var1 = var1_list
for var2 = var2_list
for i = 1 : size(example, 1)
index1 = find( (example{i, 1}(:, 2) >= var1 )==1);
index2 = find( (example{i, 1}(:, 3) >= var2 )==1);
index = intersect(index1, index2);
if ~isempty(index)
result{i, 1} = index;
end
end
final{cnt, 1} = result;
cnt = cnt + 1;
end
end
How can i do it fast way ? or How can i convert to gpuarray?
Jan on 18 Feb 2021
Ary you really sure, that you do not want to update result{i,1}, if no intersection was found?

Jan on 18 Feb 2021
Edited: Jan on 18 Feb 2021
3 times faster: Replace
index1 = find( (example{i, 1}(:, 2) >= var1 )==1);
index2 = find( (example{i, 1}(:, 3) >= var2 )==1);
index = intersect(index1, index2);
if ~isempty(index)
result{i, 1} = index;
end
by:
index1 = find(all(example{i, 1}(:, 2:3) >= [var1, var2], 2));
if ~isempty(index1)
result{i, 1} = index1;
end
And a further duplication of the speed:
index1 = (example{i, 1}(:, 2) >= var1 & ...
example{i, 1}(:, 3) >= var2);
if any(index1)
result{i, 1} = find(index1);
end
I'd expect something like
if any(index)
result{i, 1} = find(index1);
else % Reset value from former iteration?!
result{i, 1} = [];
end
but maybe this behavior is wanted. In the random tests data, this situation never happens. But if you want to set reslut{i} to [], this is faster than the IF-method above:
index1 = (example{i, 1}(:, 2) >= var1 & ...
example{i, 1}(:, 3) >= var2);
result{i} = find(index1); % Note: {i} is the same as {i, 1}
mankang on 18 Feb 2021
I saw you a lot in other people's questions.
This way is the fastest. Thanks. bro~
index1 = (example{i, 1}(:, 2) >= var1 & ...
example{i, 1}(:, 3) >= var2);
if any(index1)
result{i, 1} = find(index1);
end