Error using h5readc Unable to open the file because of HDF5 Library error. Reason:Unknown

I've just installed Matlab 2021 b to run a pre-written code and when I try to run it (the only thing I've added to the code is the file location link) Matlab gives me this error:

"Error using h5readc

Unable to open the file because of HDF5 Library error. Reason:Unknown"

related to the Matlab function code of h5read.m, line 93: [data,var_class] = h5readc(Filename,Dataset,start,count,stride);

Does anybody have any idea of what is the problem? It seems to be internal to Matlab, may it be?

Thank you,

Giulia

Answer 1

[Cris LaPierre]

If you can share your file, we can test if we get the same error. However, I would probably go straight to contacting support.

[Zeynep Berra]

Hi, I am having the same problem with the code provided on the code in this link:https://github.com/MagneticParticleImaging/MDF/blob/master/matlab/reco.m.

I deleted the download rows and placed the mdf files by myself. Also, I am using MATLAB r2023b Academic Use currently. My friend did not get the error message with r2020b, however, I want to run the code on my PC. Any thoughts?

[Cris LaPierre]

I am not able to duplicate the error on my desktop or here (see below). I recommend contacting support so they can look into it further.

%% 1. Loading the required external functions

clear all
close all

%% 2. Download measurement and systemMatrix from http://media.tuhh.de/ibi/mdf/

filenameSM = 'systemMatrix.mdf';
filenameMeas = 'measurement.mdf';

websave(filenameSM,'http://media.tuhh.de/ibi/mdfv2/systemMatrix_V2.mdf')

ans = '/users/mss.system.iyUKxB/systemMatrix.mdf'

websave(filenameMeas,'http://media.tuhh.de/ibi/mdfv2/measurement_V2.mdf')

ans = '/users/mss.system.iyUKxB/measurement.mdf'

%% 3. Loading the data
% For the System matrix (later named SM)
% to obtain infos on the file, use the command: infoSM = h5info(filename_SM);
% or read the format documentation

% read the data, saved as real numbers
S = h5read(filenameSM, '/measurement/data');

% reinterpret as complex numbers
S = complex(S.r,S.i);
% get rid of background frames
isBG = h5read(filenameSM, '/measurement/isBackgroundFrame');
S = S(isBG == 0,:,:,:);

% For the measurements
% read and convert the data as complex numbers
% note that these data contain 500 measurements
u = h5read(filenameMeas, '/measurement/data');
%u = squeeze(u(1,:,:,:) + 1i*u(2,:,:,:));
u = fft(cast(u,'double'));
u = u(1:(size(u,1)/2+1),:,:,:);

%% 4. Pre-process - Remove the frequencies which are lower than 30 kHz, as they are unreliable due to the anologue filter in the scanner

% generate frequency vector
numFreq = h5read(filenameMeas, '/acquisition/receiver/numSamplingPoints')/2+1;
rxBandwidth = h5read(filenameMeas, '/acquisition/receiver/bandwidth');
freq = linspace(0,1,numFreq) .* rxBandwidth;

% we supose that the same frequencies are measured on all channel for 
% the SM and the measurements. use only x/y receive channels
idxFreq = freq > 80e3;
S_truncated = S(:,idxFreq,1:2);
u_truncated = u(idxFreq,1:2,:);

%% 5. Merge frequency and receive channel dimensions
S_truncated = reshape(S_truncated, size(S_truncated,1), size(S_truncated,2)*size(S_truncated,3));
u_truncated = reshape(u_truncated, size(u_truncated,1)*size(u_truncated,2), size(u_truncated,3));

%% 6. Averaged the measurement used for the reconstruction over all temporal frames
u_mean_truncated = mean(u_truncated,2);

%% 7. Make two simple reconstructions
% a normalized regularized kaczmarz approach
c_normReguArt = kaczmarzReg(S_truncated(:,:),...
                        u_mean_truncated(:),...
                        1,1*10^-6,0,1,1);

% and an regularized pseudoinverse approach
[U,Sigma,V] = svd(S_truncated(:,:).','econ');
Sigma2 = diag(Sigma);
c_pseudoInverse = pseudoinverse(U,Sigma2,V,u_mean_truncated,5*10^2,1,1);

%% 8. Display an image
% read the original size of an image
number_Position = h5read(filenameSM, '/calibration/size');

figure
subplot(1,2,1)
imagesc(real(reshape(c_normReguArt(:),number_Position(1),number_Position(2))));
colormap(gray); axis square
title({'Regularized and modified ART - 3 channels';'1 iterations / lambda = 10^{-6} / real part'})

subplot(1,2,2)
imagesc(real(reshape(c_pseudoInverse(:),number_Position(1),number_Position(2))));
colormap(gray); axis square
title({'Pseudoinverse - 3 channels';' lambda = 5*10^{3} / real part'})

function [ x ] = kaczmarzReg( A,b,iterations,lambd,shuff,enforceReal,enforcePositive )
% regularized Kaczmarz
% As published here: http://iopscience.iop.org/article/10.1088/0031-9155/55/6/003/meta on page 1582.
% Other references : Saunders M 1995: Solution of sparse rectangular
% systems using LSQR and CRAIG
% or
% From Dax A 1993: On row relaxation methods for large constrained 
% least squares problems

% initialization of the variable
[N, M] = size(A);
x = complex(zeros(N,1)); 
residual = complex(zeros(M,1));
rowIndexCycle = 1:M;

% calculate the energy of each frequency component
energy = rowEnergy(A);

% may use a randomized Kaczmarz
if shuff
    rowIndexCycle = randperm(M);
end

% estimate regularization parameter
lambdZero = sum(energy.^2)/N;
lambdIter = lambd*lambdZero;

for l = 1:iterations
    for m = 1:M
        k = rowIndexCycle(m);
        
        if energy(k) > 0
            tmp = A(:,k).'*x;
            beta = (b(k) - tmp - sqrt(lambdIter)*residual(k)) / (energy(k)^2 + lambdIter);
            x = x + beta*conj(A(:,k));
            residual(k) = residual(k) + beta*sqrt(lambdIter);
        end
    end
    
    if enforceReal && ~isreal(x)
        x = complex(real(x),0);
    end
    
    if enforcePositive
        x(real(x) < 0) = 0;
    end
end

end

function [ c ] = pseudoinverse( U,Sigma,V,u,lambd,enforceReal,enforcePositive )
% This algorithm solves the Thikonov regularized least squares Problem 
% argmin(?Ax-b?� + ??b?�) using the singular value decomposition of A.
% Arguments
% `U,Sigma,V`: Compact singular value decomposition of A
% `u`: Measurement vector u
% `lambd`: The regularization parameter, relative to the matrix trace
% `enforceReal::Bool`: Enable projection of solution on real plane during iteration
% `enforcePositive::Bool`: Enable projection of solution onto positive halfplane during iteration

D = zeros(length(Sigma),1);
for i=1:length(Sigma)
    D(i) = Sigma(i)/(Sigma(i)^2+lambd^2);
end

% calculate pseudoinverse
tmp = U'*u(:); % gemv('C',...) conjugate transpose
tmp =  tmp.*D;
c = V*tmp; % gemv('N',...) not transposed

% apply constraints
if enforceReal
    c = real(c);
end
if enforcePositive
    idxNega = real(c)<0;
    c(idxNega) = 0;
end
    
end

 function [ energy ] = rowEnergy(A)		
 % Calculate the norm of each row of the input		
 % This is an image of the energy stored by each frequency component		
 % but the scaling is unclear.		
 		
 energy = sqrt(sum(abs(A.*A),1));		
 		
 end		

Answer 2

[Thomas Ulrich]

I'm also seeing this exact error in Matlab R2021b with some code that previously ran without error on older Matlab versions. Unfortunately, I also haven't figured out yet what's wrong.

[Cris LaPierre]

Same recommendation. It's hard to provide help if we don't have the code you are running and the file you are using.