how to increase the run time of the matrix ?

1 ビュー (過去 30 日間)
navanit dubey
navanit dubey 2021 年 3 月 25 日
編集済み: Jan 2021 年 3 月 25 日
Hey ,
I am runnng a code for finding the shorter path , i am able to do it easily for matrix of 37*37 but as I increase its value it start to take lot of time.I tried to preallocate few values in it but not able to do it easily.
I am attaching 2 files for the reference. less_time.m is of 37*37 matrix and more_time.m is of 148*148 matrix.
kindly guide me how to preallocate the values in it so that I can increase the time of the operations
  2 件のコメント
navanit dubey
navanit dubey 2021 年 3 月 25 日
sure sir , I added them just to get the clean values in command window.
Thankyou for the response

サインインしてコメントする。

回答 (2 件)

Matt J
Matt J 2021 年 3 月 25 日
The best option might be to use Matlab's own shortestpath routine,
  3 件のコメント
navanit dubey
navanit dubey 2021 年 3 月 25 日
yes sir but I want that in my matlab file that I have attached to it as

サインインしてコメントする。


Jan
Jan 2021 年 3 月 25 日
編集済み: Jan 2021 年 3 月 25 日
Some standard ideas for improving the runtime:
  • sqrt(X) is faster than X^0.5
  • The JIT acceleration is more powerful, if you write one command per line. Avoid constructs like this: "mink=k;minl=m;minkl=PLkm;"
  • Logical indexing is faster than numerical indexing.
% Replace:
Nonzero=find(PLK);
PLKPLK=PLK(Nonzero);
% by:
PLKPLK = PLK(PLK ~= 0);
  • This works, but the multiplication by 1 is confusing only, most of all in the combination with a lowercase L:
D = zeros(l*1,l*1);
% Nicer:
D = zeros(l, l); % But I avoid "l" consequently to reduce confusions
  • If you need 1 element only, do not waste time with searching all elements:
Select=find(Pcum>=rand);
to_visit=LJD(Select(1));
% Faster:
Select = find(Pcum>=rand, 1);
to_visit=LJD(Select);
  • In line 110 I assume you want a vector, not a matrix: "PP=zeros(Len_LJD);" Use PP=zeros(Len_LJD, 1) instead.
  • Then you can replace:
PP=zeros(Len_LJD);
for i=1:Len_LJD
PP(i)=(Tau(W,LJD(i))^Alpha)*((Eta(LJD(i)))^Beta);
end
sumpp=sum(PP);
PP=PP/sumpp;%Build probability distribution
Pcum(1)=PP(1);
for i=2:Len_LJD
Pcum(i)=Pcum(i-1)+PP(i);
end
Select=find(Pcum>=rand);
to_visit=LJD(Select(1));
% By:
PP = Tau(W,LJD).^Alpha .* Eta(LJD).^Beta;
PP = PP / sum(PP);
Pcum = cumsum(PP);
Select = find(Pcum >= rand, 1);
to_visit = LJD(Select);
  2 件のコメント
Jan
Jan 2021 年 3 月 25 日
編集済み: Jan 2021 年 3 月 25 日
Now use the profiler to find out, where the most time is spent. This is the place to optimize the code to gain more speed.

サインインしてコメントする。

Community Treasure Hunt

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

Start Hunting!

Translated by