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

Example 91 with structures._SparseFeature

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

the class _DPAdaptStruct method evaluate.

public double evaluate(_Doc doc) {
    double prob = 0, sum;
    if (m_dim == 0) {
        // not adaptation based
        for (int k = 0; k < m_cluPosterior.length; k++) {
            // need to be fixed: here we assumed binary classification
            sum = Utils.dotProduct(CLRWithDP.m_thetaStars[k].getModel(), doc.getSparse(), 0);
            if (MTCLRWithDP.m_supWeights != null && MTCLRWithDP.m_q != 0)
                sum += MTCLRWithDP.m_q * Utils.dotProduct(MTCLRWithDP.m_supWeights, doc.getSparse(), 0);
            prob += m_cluPosterior[k] * Utils.logistic(sum);
        }
    } else {
        int n, m;
        double[] As;
        for (int k = 0; k < m_cluPosterior.length; k++) {
            As = CLRWithDP.m_thetaStars[k].getModel();
            // Bias term: w_s0*a0+b0.
            sum = As[0] * CLinAdaptWithDP.m_supWeights[0] + As[m_dim];
            for (_SparseFeature fv : doc.getSparse()) {
                n = fv.getIndex() + 1;
                m = m_featureGroupMap[n];
                sum += (As[m] * CLinAdaptWithDP.m_supWeights[n] + As[m_dim + m]) * fv.getValue();
            }
            prob += m_cluPosterior[k] * Utils.logistic(sum);
        }
    }
    // accumulate the prediction results during sampling procedure
    doc.m_pCount++;
    // >0.5?1:0;
    doc.m_prob += prob;
    return prob;
}

Example 92 with structures._SparseFeature

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

the class CLRWithHDP method accumulateFeatureCount.

public void accumulateFeatureCount() {
    int cIndex = 0;
    // store the tf count of features in one cluster
    m_tf_count = new int[m_kBar][m_featureSize];
    for (_AdaptStruct user : m_userList) {
        for (_Review r : user.getReviews()) {
            if (r.getType() == rType.ADAPTATION) {
                cIndex = r.getHDPThetaStar().getIndex();
                // aggregate each cluster's word counts
                for (_SparseFeature fv : r.getSparse()) {
                    m_tf_count[cIndex][fv.getIndex()] += fv.getValue();
                }
            }
        }
    }
}

Example 93 with structures._SparseFeature

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

the class CLRWithHDP method initThetaStars.

// Randomly assign user reviews to k user groups.
@Override
public void initThetaStars() {
    initPriorG0();
    _HDPAdaptStruct user;
    double L = 0, beta_sum = Utils.sumOfArray(m_betas), betaSum_lgamma = Utils.lgamma(beta_sum), sum = 0;
    int index;
    for (_AdaptStruct u : m_userList) {
        user = (_HDPAdaptStruct) u;
        for (_Review r : user.getReviews()) {
            // for all reviews pre-compute the likelihood of being generated from a random language model
            L = 0;
            // sum = v*beta+\sum \pi_v(global language model)
            sum = beta_sum;
            // for those v with mij,v=0, frac = \gamma(beta_v)/\gamma(beta_v)=1, log frac = 0
            for (_SparseFeature fv : r.getLMSparse()) {
                index = fv.getIndex();
                sum += fv.getValue();
                // log \gamma(m_v+\pi_v+beta)/\gamma(\pi_v+beta)
                // logGamma(\beta_i) is pre-computed for efficiency
                L += Utils.lgamma(fv.getValue() + m_betas[index]) - Utils.lgamma(m_betas[index]);
            }
            L += betaSum_lgamma - Utils.lgamma(sum);
            r.setL4NewCluster(L);
            if (r.getType() == rType.TEST)
                continue;
            sampleOneInstance(user, r);
        }
    }
}

Example 94 with structures._SparseFeature

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

the class CLinAdaptWithHDP method gradientByFunc.

@Override
protected void gradientByFunc(_AdaptStruct u, _Doc review, double weight, double[] g) {
    _Review r = (_Review) review;
    // feature index
    int n, k;
    int cIndex = r.getHDPThetaStar().getIndex();
    if (cIndex < 0 || cIndex >= m_kBar)
        System.err.println("Error,cannot find the theta star!");
    int offset = m_dim * 2 * cIndex;
    double delta = (review.getYLabel() - logit(review.getSparse(), r)) * weight;
    // Bias term for individual user.
    // a[0] = ws0*x0; x0=1
    g[offset] -= delta * m_gWeights[0];
    // b[0]
    g[offset + m_dim] -= delta;
    // 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
        g[offset + k] -= delta * m_gWeights[n] * fv.getValue();
        // x_di
        g[offset + m_dim + k] -= delta * fv.getValue();
    }
}

Example 95 with structures._SparseFeature

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

the class MTCLRWithHDP method gradientByFunc.

@Override
protected void gradientByFunc(_AdaptStruct u, _Doc review, double weight, double[] g) {
    _Review r = (_Review) review;
    // feature index
    int n;
    int cIndex = r.getHDPThetaStar().getIndex();
    if (cIndex < 0 || cIndex >= m_kBar)
        System.err.println("Error,cannot find the theta star!");
    int offset = m_dim * cIndex;
    int offsetSup = m_dim * m_kBar;
    double delta = weight * (r.getYLabel() - logit(r.getSparse(), r));
    // Bias term.
    // x0=1, each cluster.
    g[offset] -= delta;
    // super model.
    g[offsetSup] -= m_q * delta;
    // Traverse all the feature dimension to calculate the gradient.
    for (_SparseFeature fv : review.getSparse()) {
        n = fv.getIndex() + 1;
        // cluster model.
        g[offset + n] -= delta * fv.getValue();
        // super model.
        g[offsetSup + n] -= delta * fv.getValue() * m_q;
    }
}