Error while calculating Cosine distance

Following is the error while calculating Cosine distance....

    Subscripted assignment dimension mismatch.
        Error in CosineDistance (line 28)
        Distance(i,j) = - (ClientMean(i,:)'*EvalSet(j,:))/(norm_x*norm_y);

 function Distance = CosineDistance(ClientMean, EvalSet)
        % Calculate Cosine Distance 
% 
%   Inputs:
%       ClientSet ---- c* dim matrix
%       EvalSet ---- n*dim matrix
%   Outputs:
%       Distance ---- c*n matrix


if ~exist('ClientMean','var')
    error('Input arguments error.');
end
if ~exist('EvalSet','var')
    error('Input arguments error.');
end

[c, d] = size(ClientMean);
[n, dim] = size(EvalSet);
if (d ~= dim)
    error('Dimensionality disagreement.');
end
for i = 1 : c
    for j = 1 : n
       
        norm_x = norm(ClientMean);
        norm_y = norm(EvalSet);
        Distance(i,j) = - (ClientMean(i,:)'*EvalSet(j,:))/(norm_x*norm_y);
    end
end

[Walter Roberson]

Are you sure about that code? It would make more sense if it was

ClientMean(i,:)*EvalSet(j,:)'

as that would be 1 x d * d x 1, giving 1 x 1.

Answer 1

[madhan ravi]

To view answer:

ClientMean=....your matrix;
EvalSet=....your matrix;
Distance = CosineDistance(ClientMean, EvalSet)  % function call
celldisp(Distance)  % after the calling of the function

Just change yours to below:

function Distance = CosineDistance(ClientMean, EvalSet)  % function definition
        % Calculate Cosine Distance 
% 
%   Inputs:
%       ClientSet ---- c* dim matrix
%       EvalSet ---- n*dim matrix
%   Outputs:
%       Distance ---- c*n matrix
if ~exist('ClientMean','var')
    error('Input arguments error.');
end
if ~exist('EvalSet','var')
    error('Input arguments error.');
end
[c, d] = size(ClientMean);
[n, dim] = size(EvalSet);
if (d ~= dim)
    error('Dimensionality disagreement.');
end

Distance=cell(c,n);  % pre-allocate
ctr=1;
for i = 1 : c
    for j = 1 : n       
        norm_x = norm(ClientMean);
        norm_y = norm(EvalSet);
        Distance{ctr} = - (ClientMean(i,:)'*EvalSet(j,:))/(norm_x*norm_y);
        ctr=ctr+1;
    end
end
end