From fe6acf199a5749f0a604a95e8d7af59bccc7505e Mon Sep 17 00:00:00 2001
From: Patrick Simianer <p@simianer.de>
Date: Fri, 9 Sep 2011 15:33:35 +0200
Subject: partial merge, ruleid feature

---
 dtrain/dtrain.cc               | 50 +++++++++++++++++++-----------------------
 dtrain/run.sh                  |  5 +++--
 dtrain/sample.h                | 18 +++++++--------
 dtrain/test/EXAMPLE/dtrain.ini |  2 +-
 4 files changed, 36 insertions(+), 39 deletions(-)

(limited to 'dtrain')

diff --git a/dtrain/dtrain.cc b/dtrain/dtrain.cc
index d58478a8..35996d6d 100644
--- a/dtrain/dtrain.cc
+++ b/dtrain/dtrain.cc
@@ -215,7 +215,7 @@ main( int argc, char** argv )
 
   // for the perceptron/SVM; TODO as params
   double eta = 0.0005;
-  double gamma = 0.01; // -> SVM
+  double gamma = 0.;//01; // -> SVM
   lambdas.add_value( FD::Convert("__bias"), 0 );
   
   // for random sampling
@@ -388,10 +388,8 @@ main( int argc, char** argv )
     if ( !noup ) {
 
       TrainingInstances pairs;
-
-      sample_all_rand(kb, pairs);
-      cout << pairs.size() << endl;
-            
+      sample_all( kb, pairs );
+       
       for ( TrainingInstances::iterator ti = pairs.begin();
             ti != pairs.end(); ti++ ) {
 
@@ -401,31 +399,29 @@ main( int argc, char** argv )
       //} else {
         //dv = ti->first - ti->second;
       //}
-        dv.add_value( FD::Convert("__bias"), -1 );
+          dv.add_value( FD::Convert("__bias"), -1 );
         
-        SparseVector<double> reg;
-        reg = lambdas * ( 2 * gamma );
-        dv -= reg;
-        lambdas += dv * eta;
-
-        if ( verbose ) {
-          cout << "{{ f("<< ti->first_rank <<") > f(" << ti->second_rank << ") but g(i)="<< ti->first_score <<" < g(j)="<< ti->second_score << " so update" << endl;
-          cout << " i  " << TD::GetString(kb->sents[ti->first_rank]) << endl;
-          cout << "    " << kb->feats[ti->first_rank] << endl;
-          cout << " j  " << TD::GetString(kb->sents[ti->second_rank]) << endl;
-          cout << "    " << kb->feats[ti->second_rank] << endl; 
-          cout << " diff vec: " << dv << endl;
-          cout << " lambdas after update: " << lambdas << endl;
-          cout << "}}" << endl;
-        }
-
+          //SparseVector<double> reg;
+          //reg = lambdas * ( 2 * gamma );
+          //dv -= reg;
+          lambdas += dv * eta;
+
+          if ( verbose ) {
+            cout << "{{ f("<< ti->first_rank <<") > f(" << ti->second_rank << ") but g(i)="<< ti->first_score <<" < g(j)="<< ti->second_score << " so update" << endl;
+            cout << " i  " << TD::GetString(kb->sents[ti->first_rank]) << endl;
+            cout << "    " << kb->feats[ti->first_rank] << endl;
+            cout << " j  " << TD::GetString(kb->sents[ti->second_rank]) << endl;
+            cout << "    " << kb->feats[ti->second_rank] << endl; 
+            cout << " diff vec: " << dv << endl;
+            cout << " lambdas after update: " << lambdas << endl;
+            cout << "}}" << endl;
+          }
         } else {
-            //if ( 0 ) {
-            SparseVector<double> reg;
-            reg = lambdas * ( gamma * 2 );
-            lambdas += reg * ( -eta );
-            //}
+          //SparseVector<double> reg;
+          //reg = lambdas * ( 2 * gamma );
+          //lambdas += reg * ( -eta );
         }
+
       }
 
       //double l2 = lambdas.l2norm();
diff --git a/dtrain/run.sh b/dtrain/run.sh
index 16575c25..97123dfa 100755
--- a/dtrain/run.sh
+++ b/dtrain/run.sh
@@ -2,9 +2,10 @@
 
 #INI=test/blunsom08.dtrain.ini
 #INI=test/nc-wmt11/dtrain.ini
-#INI=test/EXAMPLE/dtrain.ini
-INI=test/EXAMPLE/dtrain.ruleids.ini
+INI=test/EXAMPLE/dtrain.ini
+#INI=test/EXAMPLE/dtrain.ruleids.ini
 #INI=test/toy.dtrain.ini
+#INI=test/EXAMPLE/dtrain.cdecrid.ini
 
 rm /tmp/dtrain-*
 ./dtrain -c $INI $1 $2 $3 $4 
diff --git a/dtrain/sample.h b/dtrain/sample.h
index b6aa9abd..502901af 100644
--- a/dtrain/sample.h
+++ b/dtrain/sample.h
@@ -37,20 +37,20 @@ sample_all( KBestList* kb, TrainingInstances &training )
 }
 
 void
-sample_all_rand( KBestList* kb, TrainingInstances &training )
+sample_rand( KBestList* kb, TrainingInstances &training )
 {
   srand( time(NULL) );
   for ( size_t i = 0; i < kb->GetSize()-1; i++ ) {
     for ( size_t j = i+1; j < kb->GetSize(); j++ ) {
       if ( rand() % 2 ) {
-      TPair p;
-      p.first = kb->feats[i];
-      p.second = kb->feats[j];
-      p.first_rank = i;
-      p.second_rank = j;
-      p.first_score = kb->scores[i];
-      p.second_score = kb->scores[j];
-      training.push_back( p );
+        TPair p;
+        p.first = kb->feats[i];
+        p.second = kb->feats[j];
+        p.first_rank = i;
+        p.second_rank = j;
+        p.first_score = kb->scores[i];
+        p.second_score = kb->scores[j];
+        training.push_back( p );
       }
     }
   }
diff --git a/dtrain/test/EXAMPLE/dtrain.ini b/dtrain/test/EXAMPLE/dtrain.ini
index 7645921a..7221ba3f 100644
--- a/dtrain/test/EXAMPLE/dtrain.ini
+++ b/dtrain/test/EXAMPLE/dtrain.ini
@@ -5,6 +5,6 @@ epochs=3
 input=test/EXAMPLE/dtrain.nc-1k
 scorer=stupid_bleu
 output=test/EXAMPLE/weights.gz
-stop_after=100
+stop_after=1000
 wprint=Glue WordPenalty LanguageModel LanguageModel_OOV PhraseModel_0 PhraseModel_1 PhraseModel_2 PhraseModel_3 PhraseModel_4 PassThrough
 
-- 
cgit v1.2.3