mismatchstringkernel.m

来自「Proteins _ String Kernel」· M 代码 · 共 41 行

%Computes the string kernel for contiguous strings of length seqLength
%between the strings in proteinStrings.

%Arthur Gretton

function [K_mismatch]=mismatchStringKernel(seqLength,proteinStrings);

disp('Computing mismatch kernel...')

totalStringNum =length(proteinStrings);
sequenceFeatures = zeros(totalStringNum,26^seqLength);  %rows are data points


for l=1:totalStringNum
  singleString = char(proteinStrings(l));
  for m=1:length(singleString)-seqLength+1       %loop through substrings
    substring=singleString(m:m+seqLength-1);

    featureIndices = zeros(1,seqLength);     %terms that sum to the feature index
                                             %of string "substring".
    for n=1:seqLength
      featureIndices(n) = featureIndices(n)+(substring(n)-'a')*26^(n-1);
    end

    %Generate all possible feature indices which result from the
    %perturbation of ONE entry in the starting string
    for n=1:seqLength
      featureIndexBase = 1+ sum(featureIndices([1:n-1 n+1:seqLength]));
      for p=0:25
	featureIndex=featureIndexBase+p*26^(n-1);
        sequenceFeatures(l,featureIndex)=sequenceFeatures(l,featureIndex)+1;
      end
    end
  end
l;
end

K_mismatch=sequenceFeatures*sequenceFeatures';