问题描述
我最近开始使用Mathdotnet Numerics统计软件包在c#中进行数据分析.
I have recently started using Mathdotnet Numerics statistical package to do data analysis in c#.
我正在寻找互相关函数. Mathdotnet是否为此提供API?
I am looking for the cross correlation function. Does Mathdotnet have an API for this?
以前,我一直在使用MATLAB xcorr
或Python numpy.correlate
.因此,我正在寻找与这些等效的C#.
Previously I have been using MATLAB xcorr
or Python numpy.correlate
. So I am looking for a C# equivalent of these.
我仔细阅读了他们的文档,但这不是很简单. https://numerics.mathdotnet.com/api/
I have looked through their documentation but it isn't very straightforward.https://numerics.mathdotnet.com/api/
推荐答案
可以使用MathNet.Numerics.Statistics.Correlation
中的任何方法(例如Pearson
或Spearman
)来计算相关性.但是,如果要查找Matlab的xcorr
或autocorr
提供的结果,则必须使用这些方法针对输入样本之间的每个滞后/延迟值手动计算相关性.请注意,此示例同时包含交叉和自动相关.
Correlation can be calculated by any of the methods from MathNet.Numerics.Statistics.Correlation
, like Pearson
or Spearman
. But if you're looking for results like the ones provided by Matlab's xcorr
or autocorr
, then you have to manually calculate the correlation using those methods for each lag/delay value between your input samples. Notice this example includes both, cross and auto correlation.
double fs = 50; //sampling rate, Hz
double te = 1; //end time, seconds
int size = (int)(fs * te); //sample size
var t = Enumerable.Range(0, size).Select(p => p / fs).ToArray();
var y1 = t.Select(p => p < te / 2 ? 1.0 : 0).ToArray();
var y2 = t.Select(p => p < te / 2 ? 1.0 - 2*p : 0).ToArray();
var r12 = StatsHelper.CrossCorrelation(y1, y2); // Y1 * Y2
var r21 = StatsHelper.CrossCorrelation(y2, y1); // Y2 * Y1
var r11 = StatsHelper.CrossCorrelation(y1, y1); // Y1 * Y1 autocorrelation
StatsHelper:
public static class StatsHelper
{
public static LagCorr CrossCorrelation(double[] x1, double[] x2)
{
if (x1.Length != x2.Length)
throw new Exception("Samples must have same size.");
var len = x1.Length;
var len2 = 2 * len;
var len3 = 3 * len;
var s1 = new double[len3];
var s2 = new double[len3];
var cor = new double[len2];
var lag = new double[len2];
Array.Copy(x1, 0, s1, len, len);
Array.Copy(x2, 0, s2, 0, len);
for (int i = 0; i < len2; i++)
{
cor[i] = Correlation.Pearson(s1, s2);
lag[i] = i - len;
Array.Copy(s2,0,s2,1,s2.Length-1);
s2[0] = 0;
}
return new LagCorr { Corr = cor, Lag = lag };
}
}
LagCorr:
public class LagCorr
{
public double[] Lag { get; set; }
public double[] Corr { get; set; }
}
编辑:添加 Matlab 比较结果:
Adding Matlab comparison results:
clear;
step=0.02;
t=[0:step:1-step];
y1=ones(1,50);
y1(26:50)=0;
y2=[1-2*t];
y2(26:50)=0;
[cor12,lags12]=xcorr(y1,y2);
[cor21,lags21]=xcorr(y2,y1);
[cor11,lags11]=xcorr(y1,y1);
[cor22,lags22]=xcorr(y2,y2);
subplot(2,3,1);
plot(t,y1);
title('Y1');
axis([0 1 -0.5 1.5]);
subplot(2,3,2);
plot(lags12,cor12);
title('Y1*Y2');
axis([-30 30 0 15]);
subplot(2,3,3);
plot(lags11,cor11);
title('Y1*Y1');
axis([-30 30 0 30]);
subplot(2,3,4);
plot(t,y2);
title('Y2');
axis([0 1 -0.5 1.5]);
subplot(2,3,5);
plot(lags21,cor21);
title('Y2*Y1');
axis([-30 30 0 15]);
subplot(2,3,6);
plot(lags22,cor22);
title('Y2*Y2');
axis([-30 30 0 10]);
这篇关于使用Mathdotnet进行互相关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!