1 files changed, 15 insertions, 11 deletions

diff --git a/model/ranksvmtn.cpp b/model/ranksvmtn.cpp
index 1414c81..f904fdd 100644
--- a/model/ranksvmtn.cpp
+++ b/model/ranksvmtn.cpp
@@ -7,7 +7,7 @@ using namespace std;
 using namespace Eigen;
 
 // Main terminating criteria
-const int maxiter = 40; // max iteration count
+const int maxiter = 50; // max iteration count
 const double prec=1e-10; // precision
 // conjugate gradient
 const double cg_prec=1e-10; // precision
@@ -17,14 +17,21 @@ const int ls_maxiter = 10;
 const double line_prec=1e-10; // precision
 const double line_turb=1e-15; // purturbation
 
+void cal_Dw(RidList &D,const VectorXd &w, VectorXd &Dw)
+{
+    int n = D.getSize();
+    #pragma omp parallel for
+    for (int i=0;i<n;++i)
+        Dw(i) = D.getVec(i).dot(w);
+}
+
 int cal_Hs(RidList &D,const vector<int> &rank,const VectorXd &corr,const VectorXd &alpha,const VectorXd s,VectorXd &Hs)
 {
     int n = D.getSize();
     int q = D.getqSize();
     Hs = VectorXd::Zero(s.rows());
     VectorXd Ds(n);
-    for (int i=0;i<n;++i)
-        Ds(i) = D.getVec(i).dot(s);
+    cal_Dw(D,s,Ds);
     VectorXd gamma(n);
     for (int i=0;i<n;)
     {
@@ -153,8 +160,7 @@ int line_search(const VectorXd &w,RidList &D,const VectorXd &corr,const VectorXd
 {
     int n=D.getSize();
     VectorXd Dd(n);
-    for (int i=0;i<n;++i)
-        Dd(i) = D.getVec(i).dot(step);
+    cal_Dw(D,step,Dd);
     VectorXd alpha,beta,yt;
     VectorXd grad;
     VectorXd Hs;
@@ -167,8 +173,7 @@ int line_search(const VectorXd &w,RidList &D,const VectorXd &corr,const VectorXd
     while (1)
     {
         grad=w+t*step;
-        for (int i=0;i<n;++i)
-            Dd(i) = D.getVec(i).dot(grad);
+        cal_Dw(D,grad,Dd);
         cal_alpha_beta(Dd,corr,D,rank,yt,alpha,beta);
         VectorXd tmp = alpha.cwiseProduct(yt)-beta;
         VectorXd res = 0*grad;
@@ -209,8 +214,7 @@ int train_orig(int fsize, RidList &Data,const VectorXd &corr,VectorXd &weight){
     VectorXd alpha,beta;
     while (true)
     {
-        for (int i=0;i<n;++i)
-            dw(i) = Data.getVec(i).dot(weight);
+        cal_Dw(Data,weight,dw);
         cal_alpha_beta(dw,corr,Data,rank,yt,alpha,beta);
         // Generate support vector matrix sv & gradient
         obj = (weight.dot(weight) + C*(alpha.dot(yt.cwiseProduct(yt))-beta.dot(yt)))/2;
@@ -251,9 +255,9 @@ int RSVMTN::train(RidList &D){
     return 0;
 };
 
-int RSVMTN::predict(DataList &D, vector<double> &res){
+int RSVMTN::predict(RidList &D, vector<double> &res){
     res.clear();
     for (int i=0;i<D.getSize();++i)
-        res.push_back(((D.getData()[i])->feature).dot(model.weight));
+        res.push_back(D.getVec(i).dot(model.weight));
     return 0;
 };
\ No newline at end of file