Examples with structures._RankItem structures._RankItem used on opensource projects

Example 56 with structures._RankItem

use of structures._RankItem in project IR_Base by Linda-sunshine.

the class CoLinAdaptWithDiffFeatureGroups method gradientByR2.

// Calculate the gradients for the use in LBFGS.
protected void gradientByR2(_AdaptStruct user) {
    // Part 1, gradients from class 0.
    super.gradientByR2(user);
    _CoLinAdaptDiffFvGroupsStruct ui = (_CoLinAdaptDiffFvGroupsStruct) user, uj;
    int Asize = getASize(), offseti = Asize + m_dimB * 2 * ui.getId(), offsetj;
    double coef, dA, dB;
    // Part 2, gradients from class 1.
    for (_RankItem nit : ui.getNeighbors()) {
        uj = (_CoLinAdaptDiffFvGroupsStruct) m_userList.get(nit.m_index);
        offsetj = Asize + m_dimB * 2 * uj.getId();
        coef = 2 * nit.m_value;
        for (int k = 0; k < m_dimB; k++) {
            dA = coef * m_eta3 * (ui.getScalingB(k) - uj.getScalingB(k));
            dB = coef * m_eta4 * (ui.getShiftingB(k) - uj.getShiftingB(k));
            // update ui's gradient
            m_g[offseti + k] += dA;
            m_g[offseti + k + m_dimB] += dB;
            // update uj's gradient
            m_g[offsetj + k] -= dA;
            m_g[offsetj + k + m_dimB] -= dB;
        }
    }
}

Example 57 with structures._RankItem

use of structures._RankItem in project IR_Base by Linda-sunshine.

the class CoLinAdaptWithNeighborhoodLearning method calcSimA.

// Calculate the normalized similarity between a user and its neighbors.
public double[] calcSimA(_CoLinAdaptStruct ui, int i) {
    int index = 0;
    double sum = 0;
    // Normalized similarity between every pair of users.
    double[] sims = new double[ui.getNeighbors().size()];
    double[] Ai = new double[m_dim * 2];
    double[] Aj = new double[m_dim * 2];
    System.arraycopy(_CoLinAdaptStruct.sharedA, i, Ai, 0, m_dim * 2);
    for (_RankItem uj : ui.getNeighbors()) {
        System.arraycopy(_CoLinAdaptStruct.sharedA, uj.m_index, Aj, 0, m_dim * 2);
        sims[index] = Utils.cosine(Ai, Aj);
        sum += sims[index];
        index++;
    }
    // Normalize the similarities.
    for (int k = 0; k < sims.length; k++) sims[k] /= sum;
    return sims;
}

Example 58 with structures._RankItem

use of structures._RankItem in project IR_Base by Linda-sunshine.

the class CoLinAdaptWithNeighborhoodLearning method updateNeighborhood.

// Calculate the similarity with the learned weights.
public void updateNeighborhood() {
    m_diffSim = 0;
    _CoLinAdaptStruct ui;
    // Index of neighbors.
    int j = 0;
    double sim = 0;
    for (int i = 0; i < m_userList.size(); i++) {
        ui = (_CoLinAdaptStruct) m_userList.get(i);
        j = 0;
        // Traverse all neighbors.
        for (_RankItem nit : ui.getNeighbors()) {
            // New similarity between a pair of users.
            sim = logit(i, j);
            m_diffSim += (sim - nit.m_value) * (sim - nit.m_value);
            // Update similarity of neighbor.
            nit.m_value = sim;
            j++;
        }
    }
}

Example 59 with structures._RankItem

use of structures._RankItem in project IR_Base by Linda-sunshine.

the class WeightedAvgAdapt method gradientByFunc.

// shared gradient calculation by batch and online updating
@Override
protected void gradientByFunc(_AdaptStruct u, _Doc review, double weight) {
    _CoLinAdaptStruct ui = (_CoLinAdaptStruct) u;
    // general enough to accommodate both LinAdapt and CoLinAdapt
    int n, offset = m_dim * ui.getId();
    double delta = weight * (review.getYLabel() - logit(review.getSparse(), ui));
    if (m_LNormFlag)
        delta /= getAdaptationSize(ui);
    // Current user's info: Bias term + other features.
    // \theta_{ii}*x_0 and x_0=1
    m_g[offset] -= delta * ui.getSelfSim();
    for (_SparseFeature fv : review.getSparse()) {
        n = fv.getIndex() + 1;
        // \theta_{ii}*x_d
        m_g[offset + n] -= delta * ui.getSelfSim() * fv.getValue();
    }
    // Neighbors' info.
    for (_RankItem nit : ui.getNeighbors()) {
        offset = m_dim * nit.m_index;
        // neighbors' bias term.
        m_g[offset] -= delta * nit.m_value;
        for (_SparseFeature fv : review.getSparse()) {
            n = fv.getIndex() + 1;
            // neighbors' other features.
            m_g[offset + n] -= delta * nit.m_value * fv.getValue();
        }
    }
}

Example 60 with structures._RankItem

use of structures._RankItem in project IR_Base by Linda-sunshine.

the class WeightedAvgAdapt method logit.

@Override
protected // In this logit function, we need to sum over all the neighbors of the current user.
double logit(_SparseFeature[] fvs, _AdaptStruct user) {
    _CoLinAdaptStruct ui = (_CoLinAdaptStruct) user;
    // The user itself.
    double sum = ui.getSelfSim() * ui.linearFunc(fvs, 0);
    // Traverse all neighbors of the current user.
    for (_RankItem nit : ui.getNeighbors()) {
        _CoLinAdaptStruct uj = (_CoLinAdaptStruct) m_userList.get(nit.m_index);
        sum += nit.m_value * uj.linearFunc(fvs, 0);
    }
    return Utils.logistic(sum);
}