Differences

This shows you the differences between two versions of the page.

--- public:matlab_tips [2016/04/04 14:44] – [Sparse vs Dense] fangfufu
+++ public:matlab_tips [2016/04/04 14:50] – old revision restored (2016/04/02 07:41) fangfufu
@@ Line 1: / Line 1: @@
+====== Matlab Tips ======
+===== General tips =====
+  - If you plan to use parallel for-loop, put the parfor statement in the outermost loop to reduce inter-process communication.
+===== Simple text progress bar =====
+<code matlab>
+disp('     ');
+for i = 1:x
+    fprintf('\b\b\b\b\b\b%05.2f%%', i/x*100);
+    %%%%% Do your biddings here %%%%%
+end
+disp(' ');
+</code>
+===== Highly unscientific benchmark of prod() =====
+==== Sparse vs Dense ====
+<code matlab>
+>> a = rand(1000);
+>> b = eye(1000);
+>> c = sparse(a);
+>> d = sparse(b);
+>> tic; for i = 1:10000; prod(a); end; toc
+Elapsed time is 7.212577 seconds.
+>> tic; for i = 1:10000; prod(b); end; toc
+Elapsed time is 3.566560 seconds.
+>> tic; for i = 1:10000; prod(c); end; toc
+Elapsed time is 27.455137 seconds.
+>> tic; for i = 1:10000; prod(d); end; toc
+Elapsed time is 0.030041 seconds.
+>> tic; for i = 1:10000; prod(a); end; toc
+Elapsed time is 7.250969 seconds.
+>> tic; for i = 1:10000; prod(b); end; toc
+Elapsed time is 3.645549 seconds.
+>> tic; for i = 1:10000; prod(c); end; toc
+Elapsed time is 27.538796 seconds.
+>> tic; for i = 1:10000; prod(d); end; toc
+Elapsed time is 0.029214 seconds.
+>>
+>> n = 100000;
+>> a = rand(n,1);
+>> b = 1;
+>> tic; for i = 1:n; prod(a); end; toc;
+Elapsed time is 4.849304 seconds.
+>> tic; for i = 1:n; b = b * prod(a(i)); end; toc
+Elapsed time is 0.004573 seconds.
+>>
+</code>
+==== Prod vs repeated multiplication ====
+=== Script ===
+<code matlab>
+clear;
+n = 10000;
+a = rand(n,1) * exp(1);
+b = 1;
+tic;
+for i = 1:n;
+    c = prod(a);
+end;
+toc;
+tic;
+for j = 1:n
+    for i = 1:n;
+        b = b * a(i);
+    end;
+end
+toc;
+</code>
+=== Results ===
+<code matlab>
+>> test
+Elapsed time is 0.168450 seconds.
+Elapsed time is 1.003991 seconds.
+</code>
+Note that with repeated multiplication your numerical accuracy will suffer too!