RidFile tested

1 files changed, 20 insertions, 7 deletions

diff --git a/model/ranksvmtn.cpp b/model/ranksvmtn.cpp
index b82ce64..c2ca639 100644
--- a/model/ranksvmtn.cpp
+++ b/model/ranksvmtn.cpp
@@ -6,12 +6,13 @@
 using namespace std;
 using namespace Eigen;
 
-const double C=1e-2; // Compensating & scaling
+const double C=1e-5; // Compensating & scaling
 // Main terminating criteria
 const int maxiter = 10; // max iteration count
-const double prec=1e-4; // precision
+const double prec=1e-10; // precision
 // conjugate gradient
 const double cg_prec=1e-10; // precision
+const int cg_maxiter = 30;
 // line search
 const double line_prec=1e-10; // precision
 const double line_turb=1e-15; // purturbation
@@ -64,6 +65,11 @@ int cg_solve(const MatrixXd &D,const vector<int> &rank,const VectorXd &corr,cons
         x=x+p*alpha;
         res=res-q*alpha;
         ++step;
+        if (step > cg_maxiter)
+        {
+            LOG(INFO) << "CG force terminated by maxiter";
+            break;
+        }
         r_2=r_1;
     }
     return 0;
@@ -138,6 +144,7 @@ int line_search(const VectorXd &w,const MatrixXd &D,const VectorXd &corr,const v
     VectorXd grad;
     VectorXd Hs;
     vector<int> rank(D.rows());
+    int iter = 0;
 
     for (int i=0;i<A1.size();++i)
         Xd(i) = Dd(A1[i])-Dd(A2[i]);
@@ -157,6 +164,12 @@ int line_search(const VectorXd &w,const MatrixXd &D,const VectorXd &corr,const v
         t=t-g/h;
         if (g*g/h<line_prec)
             break;
+        ++iter;
+        if (iter > cg_maxiter)
+        {
+            LOG(INFO) << "line search force terminated by maxiter";
+            break;
+        }
     }
     return 0;
 }
@@ -179,14 +192,12 @@ int train_orig(int fsize, MatrixXd &D,const vector<int> &A1,const vector<int> &A
         iter+=1;
         if (iter> maxiter)
         {
-            LOG(INFO)<< "Maxiter :"<<maxiter<<" reached";
+            LOG(INFO)<< "Maxiter reached";
             break;
         }
 
         dw = D*weight;
-
         cal_alpha_beta(dw,corr,A1,A2,rank,yt,alpha,beta);
-
         // Generate support vector matrix sv & gradient
         obj = (weight.dot(weight) + C*(alpha.dot(yt.cwiseProduct(yt))-beta.dot(yt)))/2;//
         grad = weight + C*(D.transpose()*(alpha.cwiseProduct(yt)-beta));
@@ -213,8 +224,10 @@ int RSVMTN::train(DataList &D){
     vector<DataEntry*> &dat = D.getData();
     for (i=0;i<D.getSize();++i) {
         corr(i)=(dat[i])->rank>0?0.5:-0.5;
-        for (j = 0; j < D.getfSize(); ++j)
-            Data(i, j) = dat[i]->feature(j);
+
+        for (j = 0; j < D.getfSize(); ++j){
+            Data(i, j) = dat[i]->feature(j);}
+
     }
     i=j=0;
     while (i<D.getSize())