2021-07-25科研之kde_fitdist与ksdensity

matlab

pd = fitdist(x,distname)
对数据进行概率分布对象拟合/Fit probability distribution object to data
注意只限于一维 x: N*1

同ksdensity()
注意：同样限于一维

多维kde
kde toolbox,且需GCC
解决
说明文档页

bw选择方法
MISE准则
MISE准则 matlab自写code

wikipedia page中详细(见下)

n-dimensional data

---

$4.$
1-dimensional data： kernel dis方式

SixMPG = [13;15;23;29;32;34];
figure
histogram(SixMPG)

%kde拟合得到kernel distribution
pd_kernel = fitdist(data,'Kernel','BandWidth',4);
%or
pd_kernel = fitdist(x,'Kernel','Kernel','epanechnikov')
%Define the x values and compute the pdf of each distribution.
x = 50:1:250;
pdf_kernel = pdf(pd_kernel,x);
%Plot the pdf of each distribution.
plot(x, pdf_kernel, 'Color','b','LineWidth',2);
legend('Kernel Dis')
%附录：Kernel- Kernel smoother type
% normal, box, triangle, epanechnikov
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

1-dimensional data：ksdensity方式

SixMPG = [13 5 3;15 8 5;12 5 6];
x=[1 6;3 8;3 10;2 8];
x1=[11 2;7 3;1 8;9 8];
x=[1 6 9];

[f,xi] = ksdensity(SixMPG,x,'Bandwidth',4); 
[f_1,xi_1,bw_1] = ksdensity(SixMPG); 
[f_2,xi_2,bw_2] = ksdensity(SixMPG,x1);
1
2
3
4
5
6
7
8

---

$5.$
adaptive Kernel Density Estimator for High Dimensions 直接.m函数使用，无需按照mex这些鬼

记录一些快速回顾使用方式
%说明：grid:带进去compute pdf的points； Example中为了plot,所以grid费了一定功夫
%要求：data:Nd grid:Md
%pdf=akde(data,grid); % run adaptive kde

$6老师分享$
可多维kde