離散型隨機變量的聯合分布率怎么算？
給定至少兩個隨機變量X,Y,…, 它們的聯合概率分布(Joint probability distribution)指的是每一個隨機變量的值落入特定范圍或者離散點集合內的概率. 對于只有兩個隨機變量的情況, 稱為二元分布(bivariate distribution).
聯合概率分布可以使用聯合累計分布函數(joint cumulative distribution function), 連續隨機變量的聯合概率密度函數(joint probability density function)或者離散變量的聯合概率質量函數(joint probability mass function)來描述. 由此又衍生出兩個概念: 邊緣分布(marginal distribution)和條件概率分布(conditional probability distribution).
二. 離散變量的聯合概率質量函數公式
公式:
?
?是給定X=xX=x的Y=yY=y的條件概率.
而且有:
?
如果XX和YY相互獨立:
?
如果XX和YY條件不獨立(conditionally dependent):
P(X=x and Y=y)=P(X=x)?P(Y=y|X=x)P(X=x and Y=y)=P(X=x)·P(Y=y|X=x)
也可以使用聯合累計分布函數的差分來計算:
聯合累計分布函數定義是:
?
所以F(x,y)F(x,y)的導數(差分)就是P(X=x and Y=y)P(X=x and Y=y)
三. 使用Matlab計算離散2D聯合分布
參考: Calculating a 2D joint probability distribution
離散2D聯合分布可用于計算兩張圖片的互信息MI.
0. 定義兩個離散的隨機變量.
有N個點分布在邊長為1的正方形區域內. 把正方形分為K1*K2的小矩形. 統計每個小矩形內的點的個數.
% Data
N = 1e5; % number of points
xy = rand(N, 2); % coordinates of points
xy(randi(2*N, 100, 1)) = 0; % add some points on one side
xy(randi(2*N, 100, 1)) = 1; % add some points on the other side
xy(randi(N, 100, 1), :) = 0; % add some points on one corner
xy(randi(N, 100, 1), :) = 1; % add some points on one corner
inds= unique(randi(N, 100, 1));
xy(inds, :) = repmat([0 1], numel(inds), 1); % add some points on one corner
inds= unique(randi(N, 100, 1));
xy(inds, :) = repmat([1 0], numel(inds), 1); % add some points on one corner
% Intervals for rectangles
K1 = ceil(sqrt(N/5)); % number of intervals along x
K2 = K1; % number of intervals along y
int_x = [0:(1 / K1):1]; % intervals along x
int_y = [0:(1 / K2):1]; % intervals along y
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
1. 從定義出發, 使用for循環:
tic
count_cells = zeros(K1, K2);
for k1 = 1:K1
inds1 = (xy(:, 1) >= int_x(k1)) & (xy(:, 1) < int_x(k1 + 1));
for k2 = 1:K2
inds2 = (xy(:, 2) >= int_y(k2)) & (xy(:, 2) < int_y(k2 + 1));
count_cells(k1, k2) = sum(inds1 .* inds2);% 布爾相乘得到交集點的個數
end
end
toc
% Elapsed time is 39.357691 seconds.
1
2
3
4
5
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
10
11
可見使用兩重循環的計算時間非常長.
2. 使用hist3函數
N=hist3(X,'Edges',edges)是matlab中專門計算二元分布的函數.
edges是包含兩個遞增array的cell. 第一維分組edge1是edges{1}, 第二維分組edge2是edges{2}.
也就是:
edges1(i)<=X(k,1)<edges1(i+1)edges1(i)<=X(k,1)<edges1(i+1)
edges2(j)<=X(k,2)<edges2(j+1)edges2(j)<=X(k,2)<edges2(j+1)
正好落在edges1(i+1)edges1(i+1)或者edges2(j+1)edges2(j+1)上的點的個數放在N的最后一行或者最后一列.
hist3不統計edges范圍外的部分.
N是一個二維矩陣, 統計的落到每個單元格內的點的個數.
tic
count_cells_hist = hist3(xy, 'Edges', {int_x int_y});
% 注意hist3得到的矩陣是K1+1*K2+1的, 所以把最后一行和一列去掉.
% 最后一行或一列表示的是 X(k,1)= edges{1}(end)或者X(k,2) = edges{2}(end)的點數
count_cells_hist(end, :) = []; count_cells_hist(:, end) = [];
toc
all(count_cells(:) == count_cells_hist(:))
% Elapsed time is 0.017995 seconds.
1
2
3
4
5
6
7
8
1
2
3
4
5
6
7
8
顯然比用兩重for循環快多了.
3. 使用矩陣二元操作bsxfun
C = bsxfun(fun,A,B)對A和B做逐個元素的二元操作, 操作由函數 fun指定.
返回的C中, 1表示滿足條件, 0 表示不滿足條件. 可用的fun有:
fun operation
@plus Plus
@minus Minus
@timesArray multiply
@rdivideRight array divide
@ldivideLeft array divide
@power Array power
@max Binary maximum
@min Binary minimum
@rem Remainder after division
@mod Modulus after division
@atan2 Four-quadrant inverse tangent; result in radians
@atan2d Four-quadrant inverse tangent; result in degrees
@hypot Square root of sum of squares
@eq Equal
@neNot equal
@ltLess than
@le Less than or equal to
@gt Greater than
@ge Greater than or equal to
@andElement-wise logical AND
@orElement-wise logical OR
@xorLogical exclusive OR
使用bsxfun的matlab代碼:
%% bsxfun
tic
xcomps = single(bsxfun(@ge,xy(:,1),int_x));% 10000*143矩陣
ycomps = single(bsxfun(@ge,xy(:,2),int_y));% 10000*143矩陣
% 相當于求CDF
count_again = xcomps.' * ycomps; %' 143x143 = 143x1e5 * 1e5x143
% 差分后是142*142
count_again_fix = diff(diff(count_again')');
toc
% Elapsed time is 0.178316 seconds.
all(count_cells_hist(:) == count_again_fix(:))
1
2
3
4
5
6
7
8
9
10
11
1
2
3
4
5
6
7
8
9
10
11
bsxfun稍遜于hist3, 可以針對沒有statistics toolbox的情況下使用.
4. 使用accumarray
A= accumarray(subs,val)使用subs的元素值作為索引. subs和val是一一對應的. 將subs中相同值對應的val值累加. 也就是說, subs中元素的位置決定了val哪些元素相加, subs中元素的值決定了累加值在輸出中的位置. 看matlab help中示例:
Example 1
Create a 5-by-1 vector and sum values for repeated 1-D subscripts:
val = 101:105;
subs = [1; 2; 4; 2; 4];
A = accumarray(subs, val)
A =
101 % A(1) = val(1) = 101
206 % A(2) = val(2)+val(4) = 102+104 = 206
0 % A(3) = 0
208 % A(4) = val(3)+val(5) = 103+105 = 208
subs中元素值必須是正整數值. 所以在表示分組時, 可以把[0,1]區間變為[1,K1]區間. matlab代碼:
%%%%% 第五種方法Using accumarray
% Another approach is to use accumarray to make the joint histogram after we bin the data.
% Starting with int_x, int_y, K1, xy, etc.:
tic
% take (0,1) data onto [1 K1], following A.Dondas approach for easy comparison
ii = floor(xy(:,1)*(K1-eps))+1;
ii(ii<1) = 1; ii(ii>K1) = K1;
jj = floor(xy(:,2)*(K1-eps))+1;
jj(jj<1) = 1; jj(jj>K1) = K1;
% create the histogram and normalize
H = accumarray([ii jj],ones(1,size(ii,1)));
PDF = H / size(xy,1); % for probabilities summing to 1
toc
% Elapsed time is 0.006356 seconds.
all(count_cells_hist(:) == count_again_fix(:))
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
ms級別! 真是快!
5. 使用mex編譯
mex混合編程參考: 在Matlab中使用mex函數進行C/C++混合編程
#include "mex.h"
// http://stackoverflow.com/questions/19745917/calculating-a-2d-joint-probability-distribution
void mexFunction(int nlhs, mxArray *plhs[], int nrhs, const mxArray *prhs[])
{
unsigned long int hh, ctrl; /* counters */
unsigned long int N, m, n; /* size of matrices */
unsigned long int *xy; /* data */
unsigned long int *count_cells; /* joint frequencies */
/* matrices needed */
mxArray *count_cellsArray;
/* Now we need to get the data */
if (nrhs == 3) {
xy = (unsigned long int*) mxGetData(prhs[0]);
N = (unsigned long int) mxGetM(prhs[0]);//取矩陣的行數
m = (unsigned long int) mxGetScalar(prhs[1]);
n = (unsigned long int) mxGetScalar(prhs[2]);
}
/* Then build the matrices for the output */
count_cellsArray = mxCreateNumericMatrix(m + 1, n + 1, mxUINT32_CLASS, mxREAL);
count_cells = mxGetData(count_cellsArray);
plhs[0] = count_cellsArray;
hh = 0; /* counter for elements of xy */
/* for all points from 1 to N */
for(hh=0; hh<N; hh++) {
ctrl = (m + 1) * xy[N + hh] + xy[hh];
count_cells[ctrl] = count_cells[ctrl] + 1;
}
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
將代碼保存為: joint_dist_points_2D.c. 在matlab cmd中運行:
mex joint_dist_points_2D.c
1
1
生成joint_dist_points_2D.mexw32文件.
matlab調用代碼:
% Use mex function
tic
xy2 = uint32(floor(xy ./ repmat([1 / K1, 1 / K2], N, 1)));
count_cells = joint_dist_points_2D(xy2, uint32(K1), uint32(K2));
toc
% Elapsed time is 0.011696 seconds.
1
2
3
4
5
6
1
2
3
4
5
6
也是非常快的.
