Create A Program to Create Image Classification in Matlab Assignment Solution.

% Description: calculate the Euclidean distance between any two

% points

%

% Inputs:

% p: an array containing the coordinates of the first point

% q: an array containing the coordinates of the second point

%

% Outputs:

% d: a numeric value holding the straight-line distance

% between the two points

%

% Notes: you can find the equation for Euclidean distance in the

% lecture slides.

%

function d = knn_calculate_distance(p, q)

    % Guidance:

    % 1. subtract the each of the feature values in p from the

    % corresponding feature values in q

    % 2. square each of the resulting differences

    % 3. compute the total of those squared differences

    % 4. compute the square root of that total value

    % add your code and comments on the lines below:

    % Step 1

    d = q-p;

    % Step 2

    d = d.^2;

    % Step 3

    d = sum(d,2);

    % Step 4

    d = sqrt(d);

end

% Description: create a model ready to perform k-NN classification

% from some training data (assuming k=1 initially)

%

% Inputs:

% train_examples: a numeric array containing the training examples

% train_labels: a categorical array containing the associated

% labels (i.e., with the same ordering as train_examples)

%

% Outputs:

% m: a struct holding the parameters of the k-NN model (the

% training examples, the training labels, and a value for k - the number of

% nearest neighbours to use)

%

% Notes:

% This method should be relatively short and simple, perhaps just a

% few lines of code appropriately commented.

%

function m = knn_fit(train_examples, train_labels, k)

    % Guidance:

    % 1. store the supplied parameters as fields inside the model

    % struct:

    m = struct;

    % add your code and comments on the lines below:

    m.train_examples = train_examples;

    m.train_labels = train_labels;

    m.k = k;

end

% Description: use an existing k-NN model to classify some testing examples

%

% Inputs:

% m: a struct containing details of the k-NN model we want to use

% for classification

% test_examples: a numeric array containing the testing examples we want to

% classify

%

% Outputs:

% predictions: a categorical array containing the predicted

% labels (i.e., with the same ordering as test_examples)

%

% Notes:

% Assumes that the model m has been created with a call to knn_fit()

%

function predictions = knn_predict(m, test_examples)

    % Guidance (first task):

    % 1. initialise an empty categorical array to hold the predictions

    % 2. loop over every example in the testing_examples array

    % and...

    % a. find its nearest neighbour in the data inside the

    % model

    % b. take the label associated with that nearest neighbour as

    % your prediction

    % c. add the new prediction onto the end of your categorical

    % array (from step 1)

    % Guidance (second task):

    % adjust your code to take account of the k value set inside the model

    % m. Adjust step a from above so that all k nearest neighbours are

    % found. Adjust step b from above to take the the most common class

    % label across all k nearest neighbours as your prediction

    % Step 1

    predictions = categorical;

    % Step 2

    for i = 1:size(test_examples, 1)

        p = test_examples(i,:); % current example

        % Compute the distance of this example to all examples in the

        % training dataset

        d = knn_calculate_distance(m.train_examples, p);

        % Sort the distances from lowest to highest

        [d_sorted, idx] = sort(d);

        % Pick k features/label for first train example since that's the nearest neighbor

        nearest_x = m.train_examples(idx(1:m.k),:);

        nearest_y = categorical(m.train_labels(idx(1:m.k)));

        % Pick most commont feature

        mostcommon_y = mode(nearest_y);

        % Append prediction to the end of categorical array

        predictions(i) = mostcommon_y;

    end

    % add your code and comments on the lines below:

end