I'm trying to simulate 3 nearest neighbour classification without using the builtin matlab functions. I was able to 1 nearest neigbhor however i am unable to extend it to more

%% code for 1 nearest neighbor classification

total_num_test_cases = num_test_cases * 5; %75
total_num_train_cases = num_train_cases * 5; %175

% Create a vector to store assigned labels
inferredLabels = zeros(1, total_num_test_cases);

% This loop cycles through each unlabelled item:
for unlabelledCaseIdx = 1:total_num_test_cases
    unlabelledCase = testingSet(unlabelledCaseIdx, :);
    
    % As any distance is shorter than infinity
    shortestDistance = inf;
    shortestDistanceLabel = 0; % Assign a temporary label
    
    % This loop cycles through each labelled item:
    for labelledCaseIdx = 1:total_num_train_cases
        labelledCase = trainingSet(labelledCaseIdx, :);
        
        % Calculate the Euclidean distance:
        currentDist = euc(unlabelledCase, labelledCase);
        
        % Check the distance
        if currentDist < shortestDistance
            shortestDistance = currentDist;
            shortestDistanceLabel = trainingTarget(labelledCaseIdx);
        end
    end % inner loop
    
    % Assign the found label to the vector of inferred labels:
    inferredLabels(unlabelledCaseIdx) = shortestDistanceLabel
end % outer loop

Answer 1

[Image Analyst]

Attach your test and training sets of data. You can get the distances of a point to all other points in one line of code. Then sort it in ascending order.

distances = sqrt((x - allX) .^ 2 + (y - allY) .^ 2);
[sortedDistances, sortOrder] = sort(distances)
indexesOfClosest3 = sortOrder(1:3)

[Mujeeb Ul Hassan]

num_train_cases=35;
num_test_cases=length(F1)-num_train_cases;

Unrecognized function or variable 'F1'.

trainingSet=[F1(1:num_train_cases,:);
    F2(1:num_train_cases,:);
    F3(1:num_train_cases,:);
    F4(1:num_train_cases,:);
    F5(1:num_train_cases,:)]
testingSet=[F1(num_train_cases+1:end,:);
    F2(num_train_cases+1:end,:);
    F3(num_train_cases+1:end,:);
    F4(num_train_cases+1:end,:);
    F5(num_train_cases+1:end,:)]

trainingTarget=[ones(1,num_train_cases), ones(1,num_train_cases)*2, ones(1,num_train_cases)*3, ones(1,num_train_cases)*4, ones(1,num_train_cases)*5];
testingTarget=[ones(1,num_test_cases), ones(1,num_test_cases)*2, ones(1,num_test_cases)*3, ones(1,num_test_cases)*4, ones(1,num_test_cases)*5];

This is my code for test and training sets. F1 to F5 are some matrices that i have evaluated and cannot mention those because it will become complicated then.

[Image Analyst]

Did you try my suggestion?

What do you want? Do you want for every point in some array, the indexes of the 3 other points that are closest to it? So the result would be like an N-by-3 array of indexes?

[Mujeeb Ul Hassan]

Yes i want that three closest points and their arrays so that i can store them and then compare it to classify as 3 nearest neighbor

[Image Analyst]

OK so I interpret that as you don't want to try my solution yourself and just want me to do it for you. Here it is:

numPoints = 30;
allX = 100 * rand(1, numPoints);
allY = 100 * rand(1, numPoints);
plot(allX, allY, 'b.', 'MarkerSize', 40);
grid on;
hold on;
drawnow;
indexesOfClosest3 = zeros(numPoints, 3);
for k = 1 : numPoints
    x = allX(k);
    y = allY(k);
    distances = sqrt((x - allX) .^ 2 + (y - allY) .^ 2);
    [sortedDistances, sortOrder] = sort(distances);
    indexesOfClosest3(k, :) = sortOrder(2:4); % Don't include the first distance which is zero (the distance of the point to itself).
    % Plot lines to them
    for k3 = 1 : 3
        x3 = allX(indexesOfClosest3(k, k3));
        y3 = allY(indexesOfClosest3(k, k3));
        line([x, x3], [y, y3], 'Color', 'r', 'LineWidth', 2)
    end
end

Note every point is connected to the 3 other closest points. Is that what you want?

[Mujeeb Ul Hassan]

Yes, thanks alot