not enough input arguments

Iam trying to run function but it give me an error in ' d{i} = pdist2(F{i}.', testSample, 'mahalanobis');' not enough input arguments

function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, useL1distance )
% function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, useL1distance);
% 
% This function is used for classifying an unknown sample using the kNN
% algorithm, in its multi-class form.
%
% ARGUMENTS:
% - F: an CELL array that contains the feature values for each class. I.e.,
%      F{1} is a matrix of size numOfDimensions x numofSamples FOR THE FIRST
%      CLASS, etc.
%
% - testSample: the input sample to be classified
% - k: the kNN parameter
% - NORMALIZE: use class priors to weight results
% - useL1distance: use L1 instead of L2 distance
%
% RETURNS:
% - Ps: an array that contains the classification probabilities for each class
% - winnerClass: the label of the winner class

  %%error(nargchk(4,5,nargin))
  switch nargin
      case 4
           useL1distance = '1'; % euclidean distance if 4 variables included
      case 5
          useL1distance = '0'; % mahalanobis distance if 5 variables included
      otherwise
          disp('error')
  end
numOfClasses = length(F);
if (size(testSample, 2)==1)
    testSample = testSample';
end
% initilization of distance vectors:
numOfDims = zeros( 1, numOfClasses );
numOfTrainSamples = zeros( 1, numOfClasses );
d = cell(numOfClasses,1);
% d{i} is a vector, whose elements represent the distance of the testing
% sample from all the samples of i-th class
testSample(isnan(testSample)) = 0.0;
for i=1:numOfClasses
    [ numOfDims(i), numOfTrainSamples(i) ] = size( F{i} );
    d{i} = inf*ones(max(numOfTrainSamples), 1); % we fill it with inf values
    F{i}(isnan(F{i})) = 0.0;    
end
if (length(testSample)>1)
    for i=1:numOfClasses % for each class:
        if (numOfTrainSamples(i)>0)
            if ( useL1distance == 1 )                
              %  d{i} = sum( abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}'),2); % L1
                d{i} =  pdist2(F{i}.', testSample, 'euclidean');
            else
                %[size(repmat(testSample, [numOfTrainSamples(i) 1])) size(F{i}')]
                %sum(sum(isnan(F{i}')))
              %  d{i} = sum( ((repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}').^2 ),2); % L2 
                 d{i} =  pdist2(F{i}.', testSample, 'mahalanobis');   %56
            end
            d{i} = sort(d{i});            
            d{i}(end+1:max(numOfTrainSamples)) = inf;                                    
        else
            d{i} = inf;
        end        
    end
else % single dimension (NO SUM required!!!)
    for i=1:numOfClasses
        if (numOfTrainSamples(i)>0)
            d{i}   = (abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}')'); 
            d{i} = sort(d{i});
            d{i}(end+1:max(numOfTrainSamples)) = inf;            
        else
            d{i} = inf;
        end
    end    
end
kAll = zeros(numOfClasses, 1);
for j=1:k
    curArray = zeros(numOfClasses, 1);
    for i=1:numOfClasses
        curArray(i) = d{i}(kAll(i)+1);
    end    
    [MIN, IMIN] = min(curArray);
    kAll(IMIN) = kAll(IMIN) + 1;
end
if ( NORMALIZE == 0 )
    Ps = (kAll ./ k);
else
    Ps = kAll ./ numOfTrainSamples';
    Ps = Ps / sum(Ps);
end
[MAX, IMAX] = max(Ps);
winnerClass = IMAX;

>> classifyKNN_D_Multi

error

Not enough input arguments.

Error in classifyKNN_D_Multi (line 30)

numOfClasses = length(F);

this is the function Iam classifying with variable Feature is 2 class and has same column size Can anyone tell me where the problem is ?

function [label, P, classNames] = ...
    fileClassification(wavFileName, kNN, modelFileName)
%
% function [label, P, classNames] = ...
%       fileClassification(wavFileName, kNN, modelFileName)
%
% This function demonstrates the classification of an audio segment,
% stored in a wav file.
%
% ARGUMENTS:
% - wavFileName:    the path of the wav file to be classified
% - kNN:            the k parameter of the kNN algorithm
% - modelFileName:  the path of the kNN classification model
%
% RETURNS:
% - label:          the label of the winner class
% - P:              a vector that contains all estimated probabilities
%                   for each audio class contained in the model
% - classNames:     a cell array that contains the names of the 
%                   audio classes of the classification model
%
% NOTE: This function classifies the WHOLE audio file, i.e., we
%       assume that the file contains a homogeneous audio segment.
%       For mid-term classification, please use mtFileClassification().
% 
% load classification model:
[Features, classNames, MEAN, STD, Statistics, ...
    stWin, stStep, mtWin, mtStep] = kNN_model_load(modelFileName);
[x, fs] = audioread(wavFileName);         % read wav file
% short-term feature extraction:
stF = stFeatureExtraction(x, fs, stWin, stStep);      
mtWinRatio = mtWin / stWin; mtStepRatio =  mtStep / stStep;
% mid-term feature statistic calculation:
[mtFeatures] = mtFeatureExtraction(...
    stF, mtWinRatio, mtStepRatio, Statistics);
% long term averaging of the mid-term statistics:
mtFeatures = mean(mtFeatures,2);
% kNN classification
[P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN,1);

Answer 1

[Walter Roberson]

>> classifyKNN_D_Multi

You are trying to run a function that requires either 4 or 5 inputs, with no inputs at all.

      case 4
           useL1distance = '1'; % euclidean distance if 4 variables included
      case 5
          useL1distance = '0'; % mahalanobis distance if 5 variables included

The code sets useL1distance to a character regardless of what was passed in as the fifth parameter. Just the fact that you have five parameters is enough to assign the character '0' to useL1distance

   if ( useL1distance == 1 ) 

It is never going to equal the number 1 because it was assigned the character '0' or the charater '1'

[Kamil Kacer]

I thought this call will clasify using euclidean distance (4 arguments in function)

[P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN,1);

And this will use mahalanobis distance (5 arguments in function)

[P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN,1,1); 

I if doesnt work like u said what should I do instead of

if ( useL1distance == 1 ) 

something like this doeasnt work

if ( useL1distance = '1' )   or   if ( useL1distance = true

Can u please give me an example

it gives me this error when i try to classify a file

    >> [label, P, classNames] = fileClassification('cat_59.wav', 5, 'catsdogs.mat')
Matrix dimensions must agree.

Error in fileClassification (line 40)
    (mtFeatures - MEAN') ./ STD', kNN,1);

[Walter Roberson]

function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, distancetype )
% function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, distancetype);
% 
% This function is used for classifying an unknown sample using the kNN
% algorithm, in its multi-class form.
%
% ARGUMENTS:
% - F: an CELL array that contains the feature values for each class. I.e.,
%      F{1} is a matrix of size numOfDimensions x numofSamples FOR THE FIRST
%      CLASS, etc.
%
% - testSample: the input sample to be classified
% - k: the kNN parameter
% - NORMALIZE: use class priors to weight results
% - useL1distance: use L1 instead of L2 distance
%
% RETURNS:
% - Ps: an array that contains the classification probabilities for each class
% - winnerClass: the label of the winner class

  %%error(nargchk(4,5,nargin))
  switch nargin
      case 4
           distancetype = 1; % euclidean distance if 4 variables included
      case 5
          if ~isscalar(distancetype) || ~isnumeric(distancetype) || ~ismember(distancetype, 1:3);
              error('invalid distance type');
          end
      otherwise
          disp('error')
  end
numOfClasses = length(F);
if (size(testSample, 2)==1)
    testSample = testSample';
end
% initilization of distance vectors:
numOfDims = zeros( 1, numOfClasses );
numOfTrainSamples = zeros( 1, numOfClasses );
d = cell(numOfClasses,1);
% d{i} is a vector, whose elements represent the distance of the testing
% sample from all the samples of i-th class
testSample(isnan(testSample)) = 0.0;
for i=1:numOfClasses
    [ numOfDims(i), numOfTrainSamples(i) ] = size( F{i} );
    d{i} = inf*ones(max(numOfTrainSamples), 1); % we fill it with inf values
    F{i}(isnan(F{i})) = 0.0;    
end
if (length(testSample)>1)
    for i=1:numOfClasses % for each class:
        if (numOfTrainSamples(i)>0)
            if ( distancetype == 1 )                
              %  d{i} = sum( abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}'),2); % L1
                d{i} =  pdist2(F{i}.', testSample, 'euclidean');
            elseif ( distancetype == 2)
                d{i} = pdist2{F{i}.', testSample, 'cityblock');    %L1
            else
                d{i} =  pdist2(F{i}.', testSample, 'mahalanobis');   %56
            end
            d{i} = sort(d{i});            
            d{i}(end+1:max(numOfTrainSamples)) = inf;                                    
        else
            d{i} = inf;
        end        
    end
else % single dimension (NO SUM required!!!)
    for i=1:numOfClasses
        if (numOfTrainSamples(i)>0)
            d{i}   = (abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}')'); 
            d{i} = sort(d{i});
            d{i}(end+1:max(numOfTrainSamples)) = inf;            
        else
            d{i} = inf;
        end
    end    
end
kAll = zeros(numOfClasses, 1);
for j=1:k
    curArray = zeros(numOfClasses, 1);
    for i=1:numOfClasses
        curArray(i) = d{i}(kAll(i)+1);
    end    
    [MIN, IMIN] = min(curArray);
    kAll(IMIN) = kAll(IMIN) + 1;
end
if ( NORMALIZE == 0 )
    Ps = (kAll ./ k);
else
    Ps = kAll ./ numOfTrainSamples';
    Ps = Ps / sum(Ps);
end
[MAX, IMAX] = max(Ps);
winnerClass = IMAX;

Now, to use Euclidean distance use

[P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN,1); 

or

[P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN,1,1);

This is a change from before. Now the final 1 is for distance type #1, Euclidean.

To use L1 distance use distance type #2, L1

[P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN, 1, 2);

To use mahalonobis use distance type #3

[P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN, 1, 3); 

[Kamil Kacer]

Can i make it work li this ?

switch nargin
      case 4
           distancetype = 1; % euclidean distance if 4 variables included
      case 5
           distancetype = 2; 
        %  if ~isscalar(distancetype) || ~isnumeric(distancetype) || ~ismember(distancetype, 1:3)
         %     error('invalid distance type');
  
       end

and with this ?

        if ( distancetype == 1 )                
              %  d{i} = sum( abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}'),2); % L1
                d{i} =  pdist2(F{i}.', testSample, 'euclidean');
            elseif ( distancetype == 2)
                 d{i} =  pdist2(F{i}.', testSample, 'mahalanobis');   
            end

I dont get what 'cityblock' is

[Walter Roberson]

'cityblock' is another name for L1 norm.

Can i make it work li this ?

Why? If you are going to use a numeric parameter, then make it mean something.

When you go to write a GUI to allow the user to choose which distance type they want, are you going to want to code

if handles.distance_type_button.Value == 0
  [P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN, 1);
else
  [P, label] = classifyKNN_D_Multi(Features, ...
    (mtFeatures - MEAN') ./ STD', kNN, 1, 1); 
end

or are you going to want to have a single call to the function with a parameter that indicates which distance type the user wanted?

Answer 2

[Kamil Kacer]

cityblock is another form of distance?

But I do not need it.

  if ~isscalar(distancetype) || ~isnumeric(distancetype) || ~ismember(distancetype, 1:3);
              error('invalid distance type');

why is this orring everything can u please explain this ?

[Walter Roberson]

"cityblock distance" is another name for L1 norm, which is the distance measure your code used to use, the inspiration for the variable name "useL1distance" . It is also called "taxicab distance" https://en.wikipedia.org/wiki/Taxicab_geometry

There is no point in deliberately removing it -- you might want it later for further tests.

why is this orring everything

This is a common code pattern when you are trying to ensure that a variable has some specific form. Another equivalent way of writing the same code would be

  if ~(isscalar(distancetype) && isnumeric(distancetype) && ismember(distancetype, 1:3))

This is mathematically equivalent: ((~A) or (~B)) is the same as ~(A and B) . Which form you write becomes a matter of style. The ~A || ~B form is easier to add more cases to.

The tests we are doing here are:

• distancetype must be a scalar
• distancetype must be numeric
• distance type must be one of 1, 2, or 3.

[Kamil Kacer]

thank you so much

But when i try to clasificate a file it give me an error

it needs more them one row i intetionally used just one sample to train it i can understand that 'mahalanobis' may not work but i put 5 in puts in function so it suppose to do euclidean distance.

Do u know where the problem might be

fileClassification('audio.wav', 5, 'auto4.mat')
Error using pdist2 (line 296)
There must be more than one row in X to compute the default covariance matrix for Mahalanobis metric.

Error in classifyKNN_D_Multi (line 56)
                 d{i} =  pdist2(F{i}.', testSample, 'mahalanobis');

Error in fileClassification (line 43)
[P, label] = classifyKNN_D_Multi(Features, ...

[Walter Roberson]

function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, distancetype )
% function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, distancetype);
% 
% This function is used for classifying an unknown sample using the kNN
% algorithm, in its multi-class form.
%
% ARGUMENTS:
% - F: an CELL array that contains the feature values for each class. I.e.,
%      F{1} is a matrix of size numOfDimensions x numofSamples FOR THE FIRST
%      CLASS, etc.
%
% - testSample: the input sample to be classified
% - k: the kNN parameter
% - NORMALIZE: use class priors to weight results
% - useL1distance: use L1 instead of L2 distance
%
% RETURNS:
% - Ps: an array that contains the classification probabilities for each class
% - winnerClass: the label of the winner class

  %%error(nargchk(4,5,nargin))
  switch nargin
      case 4
           distancetype = 1; % euclidean distance if 4 variables included
      case 5
          distancetype = 3;
      otherwise
          disp('error')
  end
numOfClasses = length(F);
if (size(testSample, 2)==1)
    testSample = testSample';
end
% initilization of distance vectors:
numOfDims = zeros( 1, numOfClasses );
numOfTrainSamples = zeros( 1, numOfClasses );
d = cell(numOfClasses,1);
% d{i} is a vector, whose elements represent the distance of the testing
% sample from all the samples of i-th class
testSample(isnan(testSample)) = 0.0;
for i=1:numOfClasses
    [ numOfDims(i), numOfTrainSamples(i) ] = size( F{i} );
    d{i} = inf*ones(max(numOfTrainSamples), 1); % we fill it with inf values
    F{i}(isnan(F{i})) = 0.0;    
end
if (length(testSample)>1)
    for i=1:numOfClasses % for each class:
        if (numOfTrainSamples(i)>0)
            if ( distancetype == 1 )                
              %  d{i} = sum( abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}'),2); % L1
                d{i} =  pdist2(F{i}.', testSample, 'euclidean');
            elseif ( distancetype == 2)
                d{i} = pdist2{F{i}.', testSample, 'cityblock');    %L1
            else
                d{i} =  pdist2(F{i}.', testSample, 'mahalanobis');   %56
            end
            d{i} = sort(d{i});            
            d{i}(end+1:max(numOfTrainSamples)) = inf;                                    
        else
            d{i} = inf;
        end        
    end
else % single dimension (NO SUM required!!!)
    for i=1:numOfClasses
        if (numOfTrainSamples(i)>0)
            d{i}   = (abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}')'); 
            d{i} = sort(d{i});
            d{i}(end+1:max(numOfTrainSamples)) = inf;            
        else
            d{i} = inf;
        end
    end    
end
kAll = zeros(numOfClasses, 1);
for j=1:k
    curArray = zeros(numOfClasses, 1);
    for i=1:numOfClasses
        curArray(i) = d{i}(kAll(i)+1);
    end    
    [MIN, IMIN] = min(curArray);
    kAll(IMIN) = kAll(IMIN) + 1;
end
if ( NORMALIZE == 0 )
    Ps = (kAll ./ k);
else
    Ps = kAll ./ numOfTrainSamples';
    Ps = Ps / sum(Ps);
end
[MAX, IMAX] = max(Ps);
winnerClass = IMAX;

[Kamil Kacer]

fileClassification('audio.wav', 5, 'auto4.mat')
Error: File: classifyKNN_D_Multi.m Line: 52 Column: 62
Invalid expression. When calling a function or indexing a variable, use parentheses. Otherwise, check
for mismatched delimiters.

Error in fileClassification (line 43)
[P, label] = classifyKNN_D_Multi(Features, ...

There is still something wrong witrh that 'cityblock' line

[Walter Roberson]

function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, distancetype )
% function [Ps, winnerClass] = classifyKNN_D_Multi(F, testSample, k, NORMALIZE, distancetype);
% 
% This function is used for classifying an unknown sample using the kNN
% algorithm, in its multi-class form.
%
% ARGUMENTS:
% - F: an CELL array that contains the feature values for each class. I.e.,
%      F{1} is a matrix of size numOfDimensions x numofSamples FOR THE FIRST
%      CLASS, etc.
%
% - testSample: the input sample to be classified
% - k: the kNN parameter
% - NORMALIZE: use class priors to weight results
% - useL1distance: use L1 instead of L2 distance
%
% RETURNS:
% - Ps: an array that contains the classification probabilities for each class
% - winnerClass: the label of the winner class

  %%error(nargchk(4,5,nargin))
  switch nargin
      case 4
           distancetype = 1; % euclidean distance if 4 variables included
      case 5
          distancetype = 3;
      otherwise
          disp('error')
  end
numOfClasses = length(F);
if (size(testSample, 2)==1)
    testSample = testSample';
end
% initilization of distance vectors:
numOfDims = zeros( 1, numOfClasses );
numOfTrainSamples = zeros( 1, numOfClasses );
d = cell(numOfClasses,1);
% d{i} is a vector, whose elements represent the distance of the testing
% sample from all the samples of i-th class
testSample(isnan(testSample)) = 0.0;
for i=1:numOfClasses
    [ numOfDims(i), numOfTrainSamples(i) ] = size( F{i} );
    d{i} = inf*ones(max(numOfTrainSamples), 1); % we fill it with inf values
    F{i}(isnan(F{i})) = 0.0;    
end
if (length(testSample)>1)
    for i=1:numOfClasses % for each class:
        if (numOfTrainSamples(i)>0)
            if ( distancetype == 1 )                
              %  d{i} = sum( abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}'),2); % L1
                d{i} =  pdist2(F{i}.', testSample, 'euclidean');
            elseif ( distancetype == 2)
                d{i} = pdist2(F{i}.', testSample, 'cityblock');    %L1
            else
                d{i} =  pdist2(F{i}.', testSample, 'mahalanobis');   %56
            end
            d{i} = sort(d{i});            
            d{i}(end+1:max(numOfTrainSamples)) = inf;                                    
        else
            d{i} = inf;
        end        
    end
else % single dimension (NO SUM required!!!)
    for i=1:numOfClasses
        if (numOfTrainSamples(i)>0)
            d{i}   = (abs(repmat(testSample, [numOfTrainSamples(i) 1]) - F{i}')'); 
            d{i} = sort(d{i});
            d{i}(end+1:max(numOfTrainSamples)) = inf;            
        else
            d{i} = inf;
        end
    end    
end
kAll = zeros(numOfClasses, 1);
for j=1:k
    curArray = zeros(numOfClasses, 1);
    for i=1:numOfClasses
        curArray(i) = d{i}(kAll(i)+1);
    end    
    [MIN, IMIN] = min(curArray);
    kAll(IMIN) = kAll(IMIN) + 1;
end
if ( NORMALIZE == 0 )
    Ps = (kAll ./ k);
else
    Ps = kAll ./ numOfTrainSamples';
    Ps = Ps / sum(Ps);
end
[MAX, IMAX] = max(Ps);
winnerClass = IMAX;

[Kamil Kacer]

Man i just cannot make it work

>> fileClassification('audio.wav', 5, 'auto4.mat')
Error using pdist2 (line 296)
There must be more than one row in X to compute the default covariance matrix for Mahalanobis metric.

Error in classifyKNN_D_Multi (line 54)
                d{i} =  pdist2(F{i}.', testSample, 'mahalanobis');   %56

Error in fileClassification (line 43)
[P, label] = classifyKNN_D_Multi(Features, ...

[Walter Roberson]

Are you still using

[P, label] = classifyKNN_D_Multi(Features, ...
  (mtFeatures - MEAN') ./ STD', kNN,1);

[Kamil Kacer]

I had to change it just to short term features

function [label, P, classNames] = ...
    fileClassification(wavFileName, kNN, modelFileName)
%
% function [label, P, classNames] = ...
%       fileClassification(wavFileName, kNN, modelFileName)
%
% This function demonstrates the classification of an audio segment,
% stored in a wav file.
%
% ARGUMENTS:
% - wavFileName:    the path of the wav file to be classified
% - kNN:            the k parameter of the kNN algorithm
% - modelFileName:  the path of the kNN classification model
%
% RETURNS:
% - label:          the label of the winner class
% - P:              a vector that contains all estimated probabilities
%                   for each audio class contained in the model
% - classNames:     a cell array that contains the names of the 
%                   audio classes of the classification model
%
% NOTE: This function classifies the WHOLE audio file, i.e., we
%       assume that the file contains a homogeneous audio segment.
%       For mid-term classification, please use mtFileClassification().
% 
% load classification model:
 [Features, classNames, MEAN, STD, ...
    stWin, stStep] = kNN_model_load(modelFileName);


[x, fs] = audioread(wavFileName);         % read wav file
% short-term feature extraction:
ShortFeatures = stFeatureExtraction(x, fs, stWin, stStep);      
 %mtWinRatio = mtWin / stWin; mtStepRatio =  mtStep / stStep;
% mid-term feature statistic calculation:
%[mtFeatures] = mtFeatureExtraction(...
 %   stF, mtWinRatio, mtStepRatio, Statistics);
% long term averaging of the mid-term statistics:
ShortFeatures = mean(ShortFeatures,2);
% kNN classification
% % [P, label] = classifyKNN_D_Multi(Features, ...
% %     (mtFeatures - MEAN') ./ STD', kNN, 1);
[P, label] = classifyKNN_D_Multi(Features, ...
    (ShortFeatures - MEAN') ./ STD', kNN, 1,1);

[Walter Roberson]

You told me before that you were passing in only 4 arguments but were still having mahalonobis being invoked, so that is the problem that I am trying to debug. This code shows you passing in 5 arguments. You specifically asked that passing in 5 arguments should always mean mahalonobis. I argued against that but you insisted so I changed it back for you.

[Kamil Kacer]

But even with 4 arguments it gives an error

>> fileClassification('audio.wav', 5, 'auto4.mat')
Index exceeds the number of array elements (1).

Error in classifyKNN_D_Multi (line 77)
        curArray(i) = d{i}(kAll(i)+1);

Error in fileClassification (line 43)
[P, label] = classifyKNN_D_Multi(Features, ...