'too many input arguments' in mhsample

Question

Abhinav 2017-9-2

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/354927-too-many-input-arguments-in-mhsample

评论： Abhinav 2017-9-3

I am having the error 'too many input arguments' during my function call which itself is called by another built-in MatLab function 'mhsample'. I have not been able to figure out any reason for this error. Following is the snippet of the code:

% specification of unnormalized posterior pdf to sample from
 pdfPost=@(x)posteriorDist(x,C,artificialNetwork,xPrior);
% specification of proposal distribution
  pdfProp=@(x)priorDist(x,xPrior);
% specification of random number generator from proposal distribution
  generator = @(x) (xPrior(:,1)+rand(7,1).*(xPrior(:,2)-xPrior(:,1)));
% calculation of posterior of model parameters (lambda)
smpl = mhsample(x0,nsamples,'pdf',pdfPost,'proppdf',pdfProp,'proprnd',generator);

Following is the error message:

 Error using mhsample (line 117)
Error occurred while trying to evaluate the user-supplied
proppdf function '@(x)priorDist(x,xPrior)'.
 Error in metroPolisHastingSampling (line 33)
smpl =
mhsample(x0,nsamples,'pdf',pdfPost,'proppdf',pdfProp,'proprnd',generator);
 Caused by:
    Error using @(x)priorDist(x,xPrior)
    Too many input arguments.

'metroPolisHastingSampling' is the name of the script I have used for implementing Metropolis-Hastings sampling method.

So, the error is in proposal distribution which only has one input argument? So, why am I getting this error? Any useful comment is appreciated?

The code for priorDist is as follows:

 function [p] = priorDist(x,xprior)
      p=1/prod(xPrior(:,2)-xPrior(:,1));
 end

I have provided all other codes below: Here is the code for 'PosteriorDist':

 function ptr = posteriorDist(x,C,artificialNetwork)
 lh=likelihoodPost(artificialNetwork,x,C); % likelihood of observation given these parameters
 priorParam = priorDist(x); % computation of joint prior of model parameters
 ptr=lh*priorParam; % computation of unnormalized posterior
 end

rest of the definitions are

 x0=5*ones(7,1);
 nsamples=10000;
 xprior=[zeros(7,1) 15*ones(15,1)];
 C=rand(7,7);

I have taken a random just for testing purposes. artificialNetwork contains information and relevant data.

4 个评论
显示 2更早的评论隐藏 2更早的评论

Image Analyst 2017-9-2

Why are you passing x into distPrior() when you don't even use it?

Abhinav 2017-9-2

编辑：Abhinav 2017-9-2

you are right, I don't need to pass x in this case. But in a general case, I will need x too. Therefore, I kept x in priorDist function. Also, I tried removing x but it gives the same error.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2017-9-2

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/354927-too-many-input-arguments-in-mhsample#answer_280045

在 MATLAB Online 中打开

https://www.mathworks.com/help/stats/mhsample.html

"proppdf takes two arguments as inputs with the same type and size as start."

"The proposal distribution q(x,y) gives the probability density for choosing x as the next point when y is the current point. It is sometimes written as q(x|y)."

Your line

pdfProp=@(x)priorDist(x,xPrior);

should probably just be

pdfProp = @priorDist;

8 个评论
显示 6更早的评论隐藏 6更早的评论

Abhinav 2017-9-3

I understand. I will test my code again with some of your suggestions and try to find out the problem.

Abhinav 2017-9-3

Walter Roberson, I could figure out the problem finally. mhsample expects two arguments in proposal distribution function: x and y. My proposal distribution was dependent only on one argument so I passed one argument x. But when I pass two argument x and y, it works even though y is not used in my proposal distribution.

请先登录，再进行评论。