create HDF5 file for currennt

create_currennt_HDF5.m

close all;
clear all;

% load the protein dataset, this is a array of struct that has fields .pssm and .ss
% which are our features and labels respectively. pssm is a L by 20 array of features.
% ss is a L by 1 array of characters C, E, H or X which are our 4 classes.
% the actual variable that appears is called 'combined'
load protein_dataset;
% this is the name of the HDF5 file we will be writing to 
fname = ['proteins_train_currennt.nc'];

% compute the total length of all the proteins combined
N = 0;
for i = 1:length(combined)
    N = N + size(combined(i).pssm,1);
end

D=20; % the number of dimensions
input = zeros(D,N);  %our array of features (all proteins combined)
target = zeros(N,1); %our array of labels
upto = 1; % this keeps track of where we are in the input array
seqlens = []; % array of sequence lengths
for i = 1:length(combined)
    L = size(combined(i).pssm,1);
    input(:,upto:upto+L-1) = combined(i).pssm';  
    temp = combined(i).ss';
    % our labels are characters, they have to be converted to integers 0-3
    temp(temp=='C') = 0;
    temp(temp=='E') = 1;
    temp(temp=='H') = 2;
    temp(temp=='X') = 3;
    seqlens(i) = L;
    target(upto:upto+L-1) = temp;
    upto = upto + L;
end
    
% now we actually start writing the HDF5 file using MATLABS built in functionality
ncid = netcdf.create(fname,'64BIT_OFFSET');
dimid1 = netcdf.defDim(ncid,'numTimesteps',N);
dimid2 = netcdf.defDim(ncid,'inputPattSize',size(inputs,1));
dimid3 = netcdf.defDim(ncid,'numSeqs',i);
dimid4 = netcdf.defDim(ncid,'numLabels',4);
dimid5 = netcdf.defDim(ncid,'maxLabelLength',1);
dimid6 = netcdf.defDim(ncid,'maxTargStringLength',5000);
dimid7 = netcdf.defDim(ncid,'maxSeqTagLength',800);
dimid8 = netcdf.defDim(ncid,'one',1);
 
varid  = netcdf.defVar(ncid,'seqTags','char',[dimid7 dimid3]);
varid  = netcdf.defVar(ncid,'numTargetClasses','int',dimid8);
varid1 = netcdf.defVar(ncid,'inputs','double',[dimid2 dimid1]);
varid2 = netcdf.defVar(ncid,'seqLengths','int',dimid3);
varid3 = netcdf.defVar(ncid,'targetClasses','double',dimid1);
varid4 = netcdf.defVar(ncid,'labels','char',[dimid5 dimid4]);
    
netcdf.endDef(ncid);
netcdf.putVar(ncid,varid1,double(input));
netcdf.putVar(ncid,varid2,seqlens);
netcdf.putVar(ncid,varid,4);
netcdf.putVar(ncid,varid3,int32(target));
netcdf.putVar(ncid,varid4,['C';'E';'H';'X']);
netcdf.close(ncid)