PRTOOLS

cluster.m

% Loading data points
data = hall;
scatterd(hall); % 14 clusers || 3 clusters
o2c  = interactclust(data,'a');

gendats.m

data = gendats([20, 20], 2, 6); % gendats(N, D, M) --> N=length, D=dimensionality, M=distance between means
disp(size(data));
scatterd(data);
i = 10; disp(+data(i,:));

imsave.m

%% Saving NIST dataset
clc;

m = prnist([0],[1:40:1000]); % get the dataset
a  = im_box(m,0,1);
a  = im_resize(a,[28 28]);

tmp = reshape(getdata(a{1}), [28 28]); % get a single digit
disp(size(tmp))
figure; imshow(tmp)
imwrite(tmp,'image.png');

matlab_prdata.m

%% Using the PR Dataset %%

% Verift the libraries are in path
>> help prdata
>> help prdatasets
>> help prdatafiles

% random sample dataset
clc
X    = rand(10,2);
Y    = randi([0,1], 10,1);
data = prdataset(X,Y);
>> getsize(data)
>> struct(data)
>> scatterd(data, 'legend')
clc

data = kimia_images;

>> methodsview(data)  % gives the methods of an Matlab Object. In this situation the prdataset class
>> getlablist(data)   % gives you the unique set of labels 
>> getsize(data)      % gives you (num_rows, num_cols, num_classes_unique)
>> show(data(100,:))  % gives you a plot of the 100th object
>> clc

>> x = im_features(data, data, {’Area’,’Perimeter’,’Eccentricity’});
>> getsize(x)         % gives you N objects with 3 features
>> scatterd(x(1:100,:), 'legend')       % scatter plot of first 100 objects and Area wrt Perimeter
>> scatterd(x(1:100, [2,3]), 'legend')  % scatter plot of first 100 objects and Perimeter wrt Eccentricity
>> x_sub = seldat(x, [3,16]) % gives you classses - 3 and 16 from the feature set

>> data_tmp = seldat(data, [3,16]
>> getsize(data_tmp)
>> show(data_tmp)
>> struct(data_tmp)
>> data_tmp_image = data2im(data_tmp, 15); % select the 15th object from data_tmp and convert to a square matrix
>> figure; imagesc(data_tmp_image);        % plot that single image

% Image manipulations
>> data_tmp       = seldat(data, [3:3:18])
>> getsize(data_tmp)
>> data_tmp_images        = im_box(data_tmp, 0, 1) % did not quite understand this
>> figure(1); show(data_tmp_images);
>> data_tmp_images_rotate = im_rotate(data_tmp_images) % Execution is postponed until data is needed
>> figure(2); show(data_tmp_images_rotate);
>> data_tmp_image_resize  = im_resize(data_tmp_image); % Execution is postponed until data is needed
>> figure(3); show(data_tmp_image_resize);
>> showfigs;

>> preproc = im_box([],0,1)*im_rotate*im_resize([],[20 20])*im_box([],1,0);
>> data_tmp_manipulations = ddata_tmp * preproc;
>> figure; show(data_tmp_manipulations);

% Centres of gravity
>> data_tmp_2 = seldat(data_tmp, 2)
>> getsize(data_tmp_2)
>> data_tmp_2_manipulations = data_tmp_2 * preproc
>> getsize(data_tmp_2_manipulations)
>> figure; show(data_tmp_2_manipulations(1,:)); % just check what the images look like now 
>> data_tmp_2_mani_mean      = im_mean(data_tmp_2_manipulations) 
>> data_tmp_2_mani_mean_vals = double(data_tmp_2_mani_mean) % apply the conversions above
>> scatterd(data_tmp_2_mani_mean_vals)
>> figure(1); histogram(data_tmp_2_mani_mean_vals(:,1), 10);
>> figure(2); histogram(data_tmp_2_mani_mean_vals(:,2), 10);
>> showfigs;

matlab_starters.m

X = rand(10,2)
Y = randi([0,1], 10,1)
data = dataset(X,Y)

>> methodsview(data)

nonparameteric.m

% Study the histogram
>> clc
>> hist help
>> hist(rand(100),10);
>> hist(randi([-5,5],100),-5:5);

% Study PRTOOLS - Gauss
>> a = gauss(100, 0, 1); % 100 objects of mean=0 and variance=1
>> +a
>> a = gauss(10, 0); +a
>> plot(+gauss(1000,5)')
>> y = pdf('Normal', +gauss(1000,0,1)', 0, 1);
>> plot(y) %line plot and hence does not make sense
>> scatter([1:length(y)], y);

startup.m

# vim Documents/MATLAB/startup.m
%% Startup file to add packages

% PRtools
addpath('/home/__/Pattern/Code/packages/prtools')

% PRDatasets
addpath('/home/__/Pattern/Code/packages/prdatasets')
addpath('/home/__/Pattern/Code/packages/kimia')
addpath('/home/__/Pattern/Code/packages/coursedata')

% General
addpath(genpath('/home/__/Pattern/Code/packages/prtools'))

treec.m

%% Data
clc;
data          = gendats([250, 250], 2, 2);
[trn,tst]     = gendat(data,0.1); % 50 training points
fprintf('\nSize of train : %d,%d || Size of test : %d,%d', size(trn), size(tst));

%% 7.1 a)
for i=1:12
    for j=1:50
        data          = gendats([250, 250], 2, 2); % make sure the datasets aren't too far away
        [trn,tst]     = gendat(data,0.1); % 50 training points
        classifier    = treec(trn, 'infcrit', i);
        [e,c]         = testc(tst, classifier);
        fprintf('\nPruning : %d || Error : %f || Incorrect Objects : %d,%d',i, e, c);
    end
end

for i=1:12
    for j=1:1
        data          = gendats([250, 250], 2, 2); % make sure the datasets aren't too far away
        % data          = gendatb([250, 250]); % banana dataset
        % data          = gendatd([250, 250], 2, 0.5); % difficult dataset
        [trn,tst]     = gendat(data,0.1); % 50 training points
        classifier    = treec(trn, 'infcrit', i);
        [e,c]         = testc(tst, classifier);
        fprintf('\nPruning : %d || Error : %f || Incorrect Objects : %d,%d',i, e, c);
        
        figure;
        scatterd(tst);
        plotc(classifier);
        title(['Prune : ',num2str(i)]);
    end
end