Hi,
I have a table with 20 million records where I need to calculate a value which is dependent on a previous row's value and the current row's value. This new value is subsequently input to the next row. I have solved this with a FOR LOOP but with 20 million records this is very slow. I have previously got suggestions for use of cumprod and vectorized calculations which have been very useful, but due to a simplification of the original problem didn't really work. I have therefore redefind the issue and was hoping someone would have another potential solution without the FOR LOOP.
The calculation below is not a real issue, it is intended to show that I have a problem which needs to be solved sequentially with one value having been calculated is subsequently used as input into the next calculation.
Thank you for your time and help.
Kind regards,
William
B = table([1;1;1;1;1;2;2;2;3],[1;2;3;4;5;6;7;8;500]);
B.Var3 = zeros(height(B),1);
i = [false; B.Var1(1:end-1) == B.Var1(2:end)];
j = find(~i);
B.Var3(j) = B.Var2(j);
for n = 1:height(B)
if i(n) == 1
B.Var3(n) = power((sin(B.Var3(n-1)) + sqrt(B.Var2(n))) / 2,3);
end
end
>> B
B =
9×3 table
Var1 Var2 Var3
____ ____ ______
1 1 1
1 2 1.4346
1 3 2.5232
1 4 2.146
1 5 3.6351
2 6 6
2 7 1.6563
2 8 6.994
3 500 500

 Réponse acceptée

KSSV
KSSV le 22 Oct 2020
Modifié(e) : KSSV le 22 Oct 2020
B = table([1;1;1;1;1;2;2;2;3],[1;2;3;4;5;6;7;8;500]);
B.Var3 = zeros(height(B),1);
i = [false; B.Var1(1:end-1) == B.Var1(2:end)];
j = find(~i);
B.Var3(j) = B.Var2(j);
idx = find(i == 1) ;
T = sin(B.Var3) ;
B.Var3(idx) = power((T(idx-1) + sqrt(B.Var2(idx))) / 2,3);

4 commentaires

Thank you for this. Although a nice way of doing this I think the challenge is that the calculation is all done in parallel and therefor doesn't capture how a calculated value is input into the next calculation.
Values for Var3 in the original problem:
B =
9×3 table
Var1 Var2 Var3
____ ____ ________________
1 1 1
1 2 1.43464717050245
1 3 2.52322347665458
1 4 2.14595795245583
1 5 3.63511709803536
2 6 6
2 7 1.6563005366775
2 8 6.99402697270611
3 500 500
When running your suggestion
B =
9×3 table
Var1 Var2 Var3
____ ____ _________________
1 1 1
1 2 1.43464717050245
1 3 0.649519052838329
1 4 1
1 5 1.39754248593737
2 6 6
2 7 1.6563005366775
2 8 2.82842712474619
3 500 500
If one takes a look at B.Var3(3) that is the same as power(sqrt(3)/2,3) while what I wanted was power((sin(B.Var3(2)) + sqrt(B.Var2(3)))/2,3) which was a different number.
Do you see what I mean
KSSV
KSSV le 22 Oct 2020
Edited the answer....check now...
William Ambrose
William Ambrose le 22 Oct 2020
Thank you, but I still get the same answer, the issue is that I need to calculate a number and then in the next instance use the newly calculated number as input to the next calculation. Hence I see a need for sequential calculations whilst I think your approach doesn't do this, unless I am missing something.
KSSV
KSSV le 22 Oct 2020
Yes....sounds right...you are right..

Connectez-vous pour commenter.

Plus de réponses (0)

Catégories

En savoir plus sur Loops and Conditional Statements dans Centre d'aide et File Exchange

Produits

Tags

Community Treasure Hunt

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

Start Hunting!

Translated by