Examples with Classifier.supervised.modelAdaptation._AdaptStruct Classifier.supervised.modelAdaptation._AdaptStruct used on opensource projects

Example 26 with Classifier.supervised.modelAdaptation._AdaptStruct

use of Classifier.supervised.modelAdaptation._AdaptStruct in project IR_Base by Linda-sunshine.

the class LinAdapt method loadUsers.

// Initialize the weights of the transformation matrix.
@Override
public void loadUsers(ArrayList<_User> userList) {
    m_userList = new ArrayList<_AdaptStruct>();
    for (_User user : userList) m_userList.add(new _LinAdaptStruct(user, m_dim));
    m_pWeights = new double[m_gWeights.length];
}

Example 27 with Classifier.supervised.modelAdaptation._AdaptStruct

use of Classifier.supervised.modelAdaptation._AdaptStruct in project IR_Base by Linda-sunshine.

the class WeightedAvgAdapt method constructUserList.

@Override
void constructUserList(ArrayList<_User> userList) {
    int vSize = m_dim;
    // step 1: create space
    m_userList = new ArrayList<_AdaptStruct>();
    for (int i = 0; i < userList.size(); i++) {
        _User user = userList.get(i);
        // we will not create transformation matrix for this user
        m_userList.add(new _CoLinAdaptStruct(user, -1, i, m_topK));
        // Initiate user weights with global weights.
        user.setModel(m_gWeights);
    }
    m_pWeights = new double[m_gWeights.length];
    // huge space consumption
    _CoLinAdaptStruct.sharedA = new double[getVSize()];
    // step 2: copy each user's weights to shared A(weights) in _CoLinAdaptStruct
    for (int i = 0; i < m_userList.size(); i++) System.arraycopy(m_gWeights, 0, _CoLinAdaptStruct.sharedA, vSize * i, vSize);
}

Example 28 with Classifier.supervised.modelAdaptation._AdaptStruct

use of Classifier.supervised.modelAdaptation._AdaptStruct in project IR_Base by Linda-sunshine.

the class CLRWithDP method trainTrace.

// added by Lin for tracking trace.
public double trainTrace(String data, int iter) {
    m_numberOfIterations = iter;
    System.out.println(toString());
    double delta = 0, lastLikelihood = 0, curLikelihood = 0;
    int count = 0;
    // clear user performance and init cluster assignment
    init();
    // Burn in period.
    while (count++ < m_burnIn) {
        calculate_E_step();
        lastLikelihood = calculate_M_step();
    }
    try {
        String traceFile = String.format("%s_iter_%d_burnin_%d_thin_%d.txt", data, iter, m_burnIn, m_thinning);
        PrintWriter writer = new PrintWriter(new File(traceFile));
        // EM iteration.
        for (int i = 0; i < m_numberOfIterations; i++) {
            // Cluster assignment, thinning to reduce auto-correlation.
            calculate_E_step();
            // Optimize the parameters
            curLikelihood = calculate_M_step();
            delta = (lastLikelihood - curLikelihood) / curLikelihood;
            if (i % m_thinning == 0) {
                evaluateModel();
                test();
                for (_AdaptStruct u : m_userList) u.getPerfStat().clear();
            }
            writer.write(String.format("%.5f\t%.5f\t%d\t%.5f\t%.5f\n", curLikelihood, delta, m_kBar, m_perf[0], m_perf[1]));
            printInfo();
            System.out.print(String.format("[Info]Step %d: likelihood: %.4f, Delta_likelihood: %.3f\n", i, curLikelihood, delta));
            if (Math.abs(delta) < m_converge)
                break;
            lastLikelihood = curLikelihood;
        }
        writer.close();
    } catch (IOException e) {
        e.printStackTrace();
    }
    // we do not want to miss the last sample?!
    evaluateModel();
    setPersonalizedModel();
    return curLikelihood;
}

Example 29 with Classifier.supervised.modelAdaptation._AdaptStruct

use of Classifier.supervised.modelAdaptation._AdaptStruct in project IR_Base by Linda-sunshine.

the class CLRWithDP method saveClusterModels.

public void saveClusterModels(String model) {
    PrintWriter writer;
    String filename;
    File dir = new File(model);
    _thetaStar theta;
    double[] weight;
    try {
        if (!dir.exists())
            dir.mkdirs();
        for (int i = 0; i < m_kBar; i++) {
            theta = m_thetaStars[i];
            filename = String.format("%s/%d.classifier", model, theta.getIndex());
            writer = new PrintWriter(new File(filename));
            weight = theta.getModel();
            for (int v = 0; v < weight.length; v++) {
                if (v == weight.length - 1)
                    writer.write(Double.toString(weight[v]));
                else
                    writer.write(weight[v] + ",");
            }
            writer.close();
        }
        writer = new PrintWriter(new File(model + "/ClusterMember.txt"));
        for (_AdaptStruct u : m_userList) {
            _DPAdaptStruct user = (_DPAdaptStruct) u;
            writer.write(user.getUserID() + "\t" + user.getThetaStar().getIndex() + "\n");
        }
        writer.close();
    } catch (IOException e) {
        e.printStackTrace();
    }
}

Example 30 with Classifier.supervised.modelAdaptation._AdaptStruct

use of Classifier.supervised.modelAdaptation._AdaptStruct in project IR_Base by Linda-sunshine.

the class CLinAdaptWithKmeans method loadUsers.

// Initialize the weights of the transformation matrix.
@Override
public void loadUsers(ArrayList<_User> userList) {
    int totalUserSize = userList.size();
    // step 1: create space
    m_userList = new ArrayList<_AdaptStruct>();
    for (int i = 0; i < userList.size(); i++) {
        _User user = userList.get(i);
        m_userList.add(new _CLinAdaptStruct(user, m_dim, i, totalUserSize, m_clusterSize));
    }
    m_pWeights = new double[m_gWeights.length];
    // step1: init the shared A: individual + cluster + global
    _CLinAdaptStruct.sharedA = new double[getVSize()];
    for (int i = 0; i < m_userList.size() + m_clusterSize; i++) {
        for (int j = 0; j < m_dim; j++) {
            _CLinAdaptStruct.sharedA[i * m_dim * 2 + j] = 1;
        }
    }
}