用之前的非零值替换向量中的所有零 [英] Replace all zeros in vector by previous non-zero value
问题描述
Matlab/Octave 算法示例:
Matlab/Octave algorithm example:
input vector: [ 1 0 2 0 7 7 7 0 5 0 0 0 9 ]
output vector: [ 1 1 2 2 7 7 7 7 5 5 5 5 9 ]
该算法非常简单:它遍历向量并用最后一个非零值替换所有零.这看起来微不足道,当用一个缓慢的 for (i=1:length) 循环完成并且能够引用前一个元素 (i-1) 时也是如此,但看起来不可能以快速矢量化形式表达.我尝试了 merge() 和 shift() 但它只适用于第一次出现的零,而不是任意数量的零.
The algorithm is very simple: it goes through the vector and replaces all zeros with the last non-zero value. It seems trivial, and is so when done with a slow for (i=1:length) loop and being able to refer to the previous element (i-1), but looks impossible to be formulated in the fast vectorized form. I tried the merge() and shift() but it only works for the first occurrence of zero, not an arbitrary number of them.
可以在 Octave/Matlab 中以矢量化形式完成还是必须使用 C 才能在大量数据上具有足够的性能?
Can it be done in a vectorized form in Octave/Matlab or must C be used for this to have sufficient performance on big amount of data?
我有另一种类似的慢速 for-loop 算法来加速,并且在矢量化中引用以前的值似乎通常是不可能的形式,如 SQL lag()
或 group by
或 loop (i-1)
很容易做到.但是 Octave/Matlab 循环非常慢.
I have another similar slow for-loop algorithm to speed up and it seems generally impossible to refer to previous values in a vectorized form, like an SQL lag()
or group by
or loop (i-1)
would easily do. But Octave/Matlab loops are terribly slow.
有没有人找到这个普遍问题的解决方案,或者这对于基本的 Octave/Matlab 设计原因是徒劳的?
Has anyone found a solution to this general problem or is this futile for fundamental Octave/Matlab design reasons?
性能基准:
in = repmat([ 1 0 2 0 7 7 7 0 5 0 0 0 9 ] ,1 ,100000);
out = in;
tic
for i=2:length(out)
if (out(i)==0)
out(i)=out(i-1);
end
end
toc
[in(1:20); out(1:20)] % test to show side by side if ok
经过的时间是 15.047 秒.
Elapsed time is 15.047 seconds.
in = V = repmat([ 1 0 2 0 7 7 7 0 5 0 0 0 9 ] ,1 ,100000);
tic;
d = double(diff([0,V])>0);
d(find(d(2:end))+1) = find(diff([0,~V])==-1) - find(diff([0,~V])==1);
out = V(cumsum(~~V+d)-1);
toc;
[in(1:20); out(1:20)] % shows it works ok
经过的时间是 0.188167 秒.
Elapsed time is 0.188167 seconds.
15.047/0.188167 = 79.97 倍提升
15.047 / 0.188167 = 79.97 times improvement
in = repmat([ 1 0 2 0 7 7 7 0 5 0 0 0 9 ] ,1 ,100000);
a = in;
tic;
pada = [a,888];
b = pada(pada >0);
bb = b(:,1:end-1);
c = find (pada==0);
d = find(pada>0);
len = d(2:end) - (d(1:end-1));
t = accumarray(cumsum([1,len])',1);
out = bb(cumsum(t(1:end-1)));
toc;
经过的时间是 0.130558 秒.
Elapsed time is 0.130558 seconds.
15.047/0.130558 = 115.25 倍提升
15.047 / 0.130558 = 115.25 times improvement
in = repmat([ 1 0 2 0 7 7 7 0 5 0 0 0 9 ] , 1, 100000);
tic;
u = nonzeros(in);
out = u(cumsum(in~=0)).';
toc;
经过的时间是 0.0597501 秒.
Elapsed time is 0.0597501 seconds.
15.047/0.0597501 = 251.83 倍提升
15.047 / 0.0597501 = 251.83 times improvement
Slow loop: 0.010862 seconds.
Dan: 0.072561 seconds.
GameOfThrows: 0.066282 seconds.
Luis Mendo: 0.032257 seconds.
fillmissing: 0.053366 seconds.
所以我们再次得出相同的结论:MATLAB 中的循环不再慢!
So we draw yet again the same conclusion: loops in MATLAB are no longer slow!
另见:Octave/Matlab 第二部分中的微不足道/不可能的算法挑战:迭代记忆
推荐答案
以下简单的方法可以满足您的需求,而且速度可能非常快:
The following simple approach does what you want, and is probably very fast:
in = [1 0 2 0 7 7 7 0 5 0 0 0 9];
t = cumsum(in~=0);
u = nonzeros(in);
out = u(t).';
这篇关于用之前的非零值替换向量中的所有零的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!