dlgradient: covariance matrix derivative.

5 次查看（过去 30 天）
MA
MA 2021-12-13
回答： MA 2021-12-16
Assuming I have a matrix x of size (mxn), the covariance matrix is of the size nxn. I want to find the gradient of the covariance matrix with respect to the input. So, starting with this code:
function [y,dx]=cov_der(x)
y=x'*x;
dx=dlgradient(y,x,'EnableHigherDerivatives',true);
end
and evaluating it as:
[y,dx]=dlfeval(@cov_der,x)
This does not work for matrices but it works for scalars. So, is there anyway I could find the gradient with respect to every element in the matrix. THanks.

请先登录，再回答此问题。

采纳的回答

MA
MA 2021-12-16
I actually solve it. For anyone looking for the same problem:
function [y,dx]=cov_der(x)
%simple example
%x=dlarray(and(3,3));
%[y,dx]=dlfeval(@cov_der,x)
y=zeros(size(x,2),size(x,2));
z=size(x,1);
y=dlarray(y);
for i=1:size(x,2)
for j=1:size(x,2)
y(i,j)=sum(x(:,i).*x(:,j))./z;
end
end
dx=zeros(size(x,2)*size(x,2),size(x,1),size(x,2));
index=1;
for i=1:size(x,2)
for j=1:size(x,2)
dx(index,:,:)=dlgradient(y(i,j),x,'EnableHigherDerivatives',true);
index=index+1;
end
end
end

更多回答（1 个）

yanqi liu
yanqi liu 2021-12-14
yes，sir，may be use loop for every element in matrix
clc; clear all; close all;
[X1, X2] = meshgrid(linspace(0,1,10));
X1 = dlarray(X1(:));
for i = 1:length(X1)
[y(i),dx(i)]=dlfeval(@cov_der, dlarray(X1(i)));
end
% figure; plot(extractdata(X1),extractdata(y))
% hold on;
% plot(extractdata(X1),extractdata(dx))
function [y,dx]=cov_der(x)
y=x'*x;
dx=dlgradient(y,x,'EnableHigherDerivatives',true);
end
  1 个评论
MA
MA 2021-12-14
THanks for your answer, but in this example you gave the covariance is calculated for each point. I want to calculate the covariance matrix for a matrix of size mxn so the output (y) will be of size nxn.

请先登录，再进行评论。

类别

在 Help Center 和 File Exchange 中查找有关 Linear Algebra 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by