RunCode.m

function BestModel = RunCode(RNG, mode, kernel, person, X, pc,y,s)
    % This function runs the code, so it goes through and creates the model
    % for the current person
    % Parameters:
    % RNG: the random seed integer
    % mode: the activation function type, 'logistic' or 'linear'
    % kernel: whether or not there is a direct kernel transformation
    % person: the ID number of the current person
    % X: the X dataset
    % pc: the cleaned dataset vector of person IDs
    % y: the y dataset
    % s: the range of sigma to iterate over

    % Set the random seed
    rng(RNG);
    % get the data for the current person
    index                       =  pc == person;
    Xp                          =  X(index,:);
    yp                          =  y(index,1);
    index_1                     =  yp == 1;
    Xp1                         =  Xp(index_1,:);
    index_2                     =  yp == 0;
    Xp2                         =  Xp(index_2,:);
    Xp_0                        =  [Xp1;Xp2];
    Np_1                        =  size(Xp1,1);
    Np_2                        =  size(Xp2,1);
    yp_0                        =  [ones(Np_1,1);zeros(Np_2,1)];
    K                           =  [Np_1 Np_2];

    % If there's not any examples of behavior or the sample size is too
    % small, do not fit the model and return an empty one
    if or((sum(yp_0) == 0),(sum(index) < 20))
        disp('no cases of behavior or too small sample size')
        if ~exist('BestModel','var')
            BestModel.sigma     = NaN;
            BestModel.Best      = -1;
        end
        return
    end
    
    % Split into training and testing data
    [Xtrain,ytrain,Xtest,ytest] =  SplitData(Xp_0,yp_0,mode);
    [N,M]                       =  size(Xtrain);
    % Initialize variables
    RUNS                        =  1000;
    BCR_max                     =  0;



    if strcmp(kernel,'yes')
        m                       =  15;
        Prototypes              =  randsample(N,m);
        if ~exist('s','var')
            s                   = 0.2:0.005:5;
        end
        % Iterate through all possible values of sigma
        for sigma=s
            Model               =  GetADALINE_Kernel(Xtrain,ytrain,mode,sigma,m,RUNS,Prototypes);
            Model               =  EvaluateModel(Model,Xtrain,Xtest,ytest,mode,kernel,sigma,Prototypes);
            Model.Prototypes    =  Prototypes;
            fprintf('sigma = %.3f, Accuracy(train) = %.4f and Accuracy(test) = %.4f\n',sigma,Model.ACCtr,Model.BCR);
            if Model.BCR > BCR_max
                BestModel       =  Model;
                BestModel.sigma =  sigma;
                BCR_max         =  Model.BCR;
            end
        end

    elseif strcmp(kernel,'no')
        Model                   =  GetADALINE(Xtrain,ytrain,mode,RUNS);
        Model                   =  EvaluateModel(Model,Xtrain,Xtest,ytest,mode,kernel,[],[]);
    end
    


    % If the best model cannot be found, fill it with empty fields
    if ~exist('BestModel','var')
        BestModel               = Model;
        if ~exist('sigma','var')
            BestModel.sigma     = NaN;
        else
            BestModel.sigma     = sigma;
        end
        BestModel.Best          = false;
    else % if the best model can be found, indicate it as such
        BestModel.Best          = true;
    end
    % Save training and test sets in the model
    BestModel.Xtrain            =  Xtrain;
    BestModel.ytrain            =  ytrain;
    BestModel.Xtest             =  Xtest;
    BestModel.ytest             =  ytest;
end