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

Example 81 with structures._SparseFeature

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

the class LinAdapt method linearFunc.

protected double linearFunc(_SparseFeature[] fvs, _AdaptStruct u) {
    _LinAdaptStruct user = (_LinAdaptStruct) u;
    // Bias term: w0*a0+b0.
    double value = user.getScaling(0) * m_gWeights[0] + user.getShifting(0);
    // feature index and feature group index
    int n = 0, k = 0;
    for (_SparseFeature fv : fvs) {
        n = fv.getIndex() + 1;
        k = m_featureGroupMap[n];
        value += (user.getScaling(k) * m_gWeights[n] + user.getShifting(k)) * fv.getValue();
    }
    return value;
}

Example 82 with structures._SparseFeature

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

the class MTLinAdapt method logit.

// We can do A_i*A_s*w_g*x at the same time to reduce computation.
@Override
protected double logit(_SparseFeature[] fvs, _AdaptStruct u) {
    // feature index and feature group index
    int n = 0, k = 0;
    _CoLinAdaptStruct ui = (_CoLinAdaptStruct) u;
    // Bias term: w_s0*a0+b0.
    double value = ui.getScaling(0) * getSupWeights(0) + ui.getShifting(0);
    for (_SparseFeature fv : fvs) {
        n = fv.getIndex() + 1;
        k = m_featureGroupMap[n];
        value += (ui.getScaling(k) * getSupWeights(n) + ui.getShifting(k)) * fv.getValue();
    }
    return 1 / (1 + Math.exp(-value));
}

Example 83 with structures._SparseFeature

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

the class MTLinAdapt method gradientByFunc.

// Gradients from loglikelihood, contributes to both individual user's gradients and super user's gradients.
protected void gradientByFunc(_AdaptStruct u, _Doc review, double weight) {
    _CoLinAdaptStruct ui = (_CoLinAdaptStruct) u;
    // feature index and feature group index
    int n, k, s;
    // general enough to accommodate both LinAdapt and CoLinAdapt
    int offset = 2 * m_dim * ui.getId();
    int offsetSup = 2 * m_dim * m_userList.size();
    double delta = weight * (review.getYLabel() - logit(review.getSparse(), ui));
    if (m_LNormFlag)
        delta /= getAdaptationSize(ui);
    // Bias term for individual user.
    // a[0] = ws0*x0; x0=1
    m_g[offset] -= delta * getSupWeights(0);
    // b[0]
    m_g[offset + m_dim] -= delta;
    // Bias term for super user.
    // a_s[0] = a_i0*w_g0*x_d0
    m_g[offsetSup] -= delta * ui.getScaling(0) * m_gWeights[0];
    // b_s[0] = a_i0*x_d0
    m_g[offsetSup + m_dimSup] -= delta * ui.getScaling(0);
    // Traverse all the feature dimension to calculate the gradient for both individual users and super user.
    for (_SparseFeature fv : review.getSparse()) {
        n = fv.getIndex() + 1;
        k = m_featureGroupMap[n];
        // w_si*x_di
        m_g[offset + k] -= delta * getSupWeights(n) * fv.getValue();
        // x_di
        m_g[offset + m_dim + k] -= delta * fv.getValue();
        s = m_featureGroupMap4SupUsr[n];
        // a_i*w_gi*x_di
        m_g[offsetSup + s] -= delta * ui.getScaling(k) * m_gWeights[n] * fv.getValue();
        // a_i*x_di
        m_g[offsetSup + m_dimSup + s] -= delta * ui.getScaling(k) * fv.getValue();
    }
}

Example 84 with structures._SparseFeature

use of structures._SparseFeature 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 85 with structures._SparseFeature

use of structures._SparseFeature 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);
}