Using a 2 dimensional feature

I am current student and new to Matlab. For an assignment I was told that I was not using the 2 dimensional feature for a .m file that I loaded. How do I do this and how do I figure it out? Where are my mistakes?

partTwoData is a 10000x2 double.

My variables made are coming up as x1

load ('partTwoData.mat');

numberOfClassOneSamples = size(classOne, 1);
numberOfClassTwoSamples = size(classTwo, 1);

%Generate a random partition of the data, 
randomizedIndices = randsample(numberOfClassOneSamples, numberOfClassOneSamples);

%splitting each of the classes into 60% training and 40%    
classOneTrainingIndices = randomizedIndices(1:6000);
classOneTestIndices= randomizedIndices(6001:10000); 
%for classTwo
classTwoTrainingIndices = randomizedIndices(1:6000);
classTwoTestIndices= randomizedIndices(6001:10000); 

classOneTrainingData = classOne(classOneTrainingIndices);
classOneTestData = classOne(classOneTestIndices); 

classTwoTrainingData = classTwo(classTwoTrainingIndices);
classTwoTestData = classTwo(classTwoTestIndices);

numberOfTestSamples = 4000; 
numberOfTrainingSamples = 6000;

%class one test samples
for i=1:numberOfTestSamples
    for j=1:numberOfTrainingSamples
        euclideanDistanceClassOne(j) = sqrt ( (classOneTestData(i) - classOneTrainingData(j))^2);
    end
    
    for j=1 : numberOfTrainingSamples
        euclideanDistanceClassTwo(j) = sqrt( (classOneTestData(i) - classTwoTrainingData(j))^2);
    end
    
    if(min(euclideanDistanceClassOne) < min(euclideanDistanceClassTwo))
        %classify as class one
        predict(i) = 1;
    else
        predict(i) = 2;
    end
    
end

correctClassOne = sum(predict==1);
incorrectClassOne = sum(predict==2);

%Class Two Test Samples, still j, because we are stepping through the training sample. 
for i=1:numberOfTestSamples 
    for j=1:numberOfTrainingSamples
         euclideanDistanceClassOne(j) = sqrt((classTwoTestData(i) - classOneTrainingData(j))^2);
    end 
    
    for j=1:numberOfTrainingSamples
       euclideanDistanceClassTwo(j) = sqrt((classTwoTestData(i) - classTwoTrainingData(j))^2);
    end 
    
    if (min(euclideanDistanceClassOne) < min(euclideanDistanceClassTwo))
        %classify as class one
        predict(i)=1; 
    else 
        %classify as class two
        predict(i)=2;
    end 
end 

correctClassTwo = sum(predict == 2);
incorrectClassTwo = sum(predict == 1);

fprintf('Part Two:');
fprintf('Class One Correct Predictions = %d\n', correctClassOne);
fprintf('Class Two Correct Predictions = %d\n', correctClassTwo);

fprintf('Class One Incorrect Predictions = %d\n', incorrectClassOne);
fprintf('Class Two Incorrect Predictions = %d\n', incorrectClassTwo);

fprintf('Total Class One Incorrect Predictions = %d\n', correctClassOne + correctClassTwo);
fprintf('Total Class Two Incorrect Predictions = %d\n', incorrectClassOne + incorrectClassTwo);
fprintf('Total Prediction Accuracy = %f\n', (correctClassOne + correctClassTwo) / (2 * numberOfTestSamples));

Answer 1

[Walter Roberson]

Every place that you index classTwoTestData(some_index) you need to change it to classTwoTestData(some_index,:) and likewise for similar variables.

You also need to adjust your euclidean distance calculations. A(index,:) - B(index,:) will be a vector, so you cannot use ^2 for it: you need to use .^2 . You also need to take the sum of the squared components.

[G. Nardi]

Thank you! Can you better explain the adjustment needed for the eucledian distance?

I got this error

>>

Unable to perform assignment because the left and right sides have a different number of

elements.

Error in Week2pt2 (line 29)

euclideanDistanceClassOne(j) = sqrt ( (classOneTestData(i,:) -

classOneTrainingData(j,:)).^2);

load ('partTwoData.mat');

numberOfClassOneSamples = size(classOne, 1);
numberOfClassTwoSamples = size(classTwo, 1);

%Generate a random partition of the data, 
randomizedIndices = randsample(numberOfClassOneSamples, numberOfClassOneSamples);

%splitting each of the classes into 60% training and 40%    
classOneTrainingIndices = randomizedIndices(1:6000,:);
classOneTestIndices= randomizedIndices(6001:10000,:); 
%for classTwo
classTwoTrainingIndices = randomizedIndices(1:6000,:);
classTwoTestIndices= randomizedIndices(6001:10000,:); 

classOneTrainingData = classOne(classOneTrainingIndices,:);
classOneTestData = classOne(classOneTestIndices,:); 

classTwoTrainingData = classTwo(classTwoTrainingIndices,:);
classTwoTestData = classTwo(classTwoTestIndices,:);

numberOfTestSamples = 4000; 
numberOfTrainingSamples = 6000;

%class one test samples
for i=1:numberOfTestSamples
    for j=1:numberOfTrainingSamples
        euclideanDistanceClassOne(j) = sqrt ( (classOneTestData(i,:) - classOneTrainingData(j,:)).^2);
    end
    
    for j=1 : numberOfTrainingSamples
        euclideanDistanceClassTwo(j) = sqrt( (classOneTestData(i,:) - classTwoTrainingData(j,:)).^2);
    end
    
    if(min(euclideanDistanceClassOne) < min(euclideanDistanceClassTwo))
        %classify as class one
        predict(i) = 1;
    else
        predict(i) = 2;
    end
    
end

correctClassOne = sum(predict==1);
incorrectClassOne = sum(predict==2);

%Class Two Test Samples, still j, because we are stepping through the training sample. 
for i=1:numberOfTestSamples 
    for j=1:numberOfTrainingSamples
         euclideanDistanceClassOne(j) = sqrt((classTwoTestData(i) - classOneTrainingData(j))^2);
    end 
    
    for j=1:numberOfTrainingSamples
       euclideanDistanceClassTwo(j) = sqrt((classTwoTestData(i) - classTwoTrainingData(j))^2);
    end 
    
    if (min(euclideanDistanceClassOne) < min(euclideanDistanceClassTwo))
        %classify as class one
        predict(i)=1; 
    else 
        %classify as class two
        predict(i)=2;
    end 
end 

correctClassTwo = sum(predict == 2);
incorrectClassTwo = sum(predict == 1);

fprintf('Part Two:');
fprintf('Class One Correct Predictions = %d\n', correctClassOne);
fprintf('Class Two Correct Predictions = %d\n', correctClassTwo);

fprintf('Class One Incorrect Predictions = %d\n', incorrectClassOne);
fprintf('Class Two Incorrect Predictions = %d\n', incorrectClassTwo);

fprintf('Total Class One Incorrect Predictions = %d\n', correctClassOne + correctClassTwo);
fprintf('Total Class Two Incorrect Predictions = %d\n', incorrectClassOne + incorrectClassTwo);
fprintf('Total Prediction Accuracy = %f\n', (correctClassOne + correctClassTwo) / (2 * numberOfTestSamples));

[Walter Roberson]

Consider that 2d Euclidean is

sqrt( (x2-x1).^2 + (y2-y1).^2 )

and consider that your existing code is

sqrt( [(x2-x1).^2, (y2-y1).^2] )

it should be obvious what is missing .

[G. Nardi]

I am still a student and trying to learn. Nothing is really obvious.

I tried

%class one test samples
for i=1:numberOfTestSamples
    for j=1:numberOfTrainingSamples
        euclideanDistanceClassOne(j) = sqrt ( (classOneTestData(i,:) - classOneTrainingData(j,:)).^2);
    end
    

But I get this error

Unable to perform assignment because the left and right sides have a different number of

elements.

Error in

euclideanDistanceClassOne(j) = sqrt ( (classOneTestData(i,:) -

classOneTrainingData(j,:)).^2);

[Walter Roberson]

If you have a vector of values, such as

(classOneTestData(i,:) - classOneTrainingData(j,:)).^2

is a vector of values, then how would you find the total value?

[G. Nardi]

sum! got it thanks.