test now passing

18 files changed, 126 insertions, 135 deletions

diff --git a/Makefile b/Makefile
index 910adad0..0d53bb54 100644
--- a/Makefile
+++ b/Makefile
@@ -65,7 +65,7 @@ FLAGS = -std=c++0x $(CFLAGS) $(LDFLAGS) $(ARCH) $(WARN_FLAGS) $(OPTIM_FLAGS) -D_
 #CXX = g++
 
 # for valgrind / gdb debugging
-FLAGS = -std=c++0x $(CFLAGS) $(LDFLAGS) $(ARCH) $(WARN_FLAGS) -ffast-math -D_FILE_OFFSET_BITS=64 $(BOOST_INCLUDE) -g -O0  -fPIC
+#FLAGS = -std=c++0x $(CFLAGS) $(LDFLAGS) $(ARCH) $(WARN_FLAGS) -ffast-math -D_FILE_OFFSET_BITS=64 $(BOOST_INCLUDE) -g -O0  -fPIC
 
 # for valgrind profiling: run 'valgrind --tool=callgrind PROGRAM' then 'callgrind_annotate --tree=both --inclusive=yes'
 #FLAGS = $(CFLAGS) $(LDFLAGS) -Wall $(ARCH) -ffast-math -D_FILE_OFFSET_BITS=64 $(BOOST_INCLUDE) -g -O2 -fomit-frame-pointer -ffast-math -fno-strict-aliasing  -fPIC
diff --git a/library/gd_mf_weights.cc b/library/gd_mf_weights.cc
index 4b394775..268739a4 100644
--- a/library/gd_mf_weights.cc
+++ b/library/gd_mf_weights.cc
@@ -52,12 +52,17 @@ int main(int argc, char *argv[])
   vw* model = VW::initialize(vwparams);

   model->audit = true;

 

+  string target("--rank ");

+  size_t loc = vwparams.find(target);

+  const char* location = vwparams.c_str()+loc+target.size();

+  size_t rank = atoi(location);

+  

   // global model params

   unsigned char left_ns = model->pairs[0][0];

   unsigned char right_ns = model->pairs[0][1];

   weight* weights = model->reg.weight_vector;

   size_t mask = model->reg.weight_mask;

-

+  

   // const char *filename = argv[0];

   FILE* file = fopen(infile.c_str(), "r");

   char* line = NULL;

@@ -86,7 +91,7 @@ int main(int argc, char *argv[])
 	      left_linear << f->feature << '\t' << weights[f->weight_index & mask];

 

 	      left_quadratic << f->feature;

-	      for (size_t k = 1; k <= model->rank; k++)

+	      for (size_t k = 1; k <= rank; k++)

 		left_quadratic << '\t' << weights[(f->weight_index + k) & mask];

 	    }

 	  left_linear << endl;

@@ -101,8 +106,8 @@ int main(int argc, char *argv[])
 	      right_linear << f->feature << '\t' << weights[f->weight_index & mask];

 

 	      right_quadratic << f->feature;

-	      for (size_t k = 1; k <= model->rank; k++)

-		right_quadratic << '\t' << weights[(f->weight_index + k + model->rank) & mask];

+	      for (size_t k = 1; k <= rank; k++)

+		right_quadratic << '\t' << weights[(f->weight_index + k + rank) & mask];

 	    }

 	  right_linear << endl;

 	  right_quadratic << endl;

diff --git a/test/test-sets/ref/ml100k_small.stderr b/test/test-sets/ref/ml100k_small.stderr
index c8bcbb51..92b91e64 100644
--- a/test/test-sets/ref/ml100k_small.stderr
+++ b/test/test-sets/ref/ml100k_small.stderr
@@ -3,7 +3,6 @@ Num weight bits = 16
 learning rate = 10
 initial_t = 1
 power_t = 0.5
-rank = 10
 using no cache
 Reading datafile = test-sets/ml100k_small_test
 num sources = 1
diff --git a/test/train-sets/ref/ml100k_small.stderr b/test/train-sets/ref/ml100k_small.stderr
index 3f3d6e25..462dca37 100644
--- a/test/train-sets/ref/ml100k_small.stderr
+++ b/test/train-sets/ref/ml100k_small.stderr
@@ -6,7 +6,6 @@ learning rate = 0.05
 initial_t = 1
 power_t = 0
 decay_learning_rate = 0.97
-rank = 10
 creating cache_file = train-sets/ml100k_small_train.cache
 Reading datafile = train-sets/ml100k_small_train
 num sources = 1
diff --git a/test/train-sets/ref/sequencespan_data.nonldf-bilou.test.stderr b/test/train-sets/ref/sequencespan_data.nonldf-bilou.test.stderr
index aad69ced..5b3afa22 100644
--- a/test/train-sets/ref/sequencespan_data.nonldf-bilou.test.stderr
+++ b/test/train-sets/ref/sequencespan_data.nonldf-bilou.test.stderr
@@ -1,10 +1,10 @@
 only testing
+switching to BILOU encoding for sequence span labeling
 Num weight bits = 18
 learning rate = 10
 initial_t = 1
 power_t = 0.5
 predictions = sequencespan_data.predict
-switching to BILOU encoding for sequence span labeling
 using no cache
 Reading datafile = train-sets/sequencespan_data
 num sources = 1
diff --git a/test/train-sets/ref/sequencespan_data.nonldf-bilou.train.stderr b/test/train-sets/ref/sequencespan_data.nonldf-bilou.train.stderr
index f0c914e0..a4fc72a6 100644
--- a/test/train-sets/ref/sequencespan_data.nonldf-bilou.train.stderr
+++ b/test/train-sets/ref/sequencespan_data.nonldf-bilou.train.stderr
@@ -1,10 +1,10 @@
 final_regressor = models/sequencespan_data.model
+switching to BILOU encoding for sequence span labeling
 Num weight bits = 18
 learning rate = 10
 initial_t = 1
 power_t = 0.5
 decay_learning_rate = 1
-switching to BILOU encoding for sequence span labeling
 creating cache_file = train-sets/sequencespan_data.cache
 Reading datafile = train-sets/sequencespan_data
 num sources = 1
diff --git a/test/train-sets/ref/sequencespan_data.nonldf.train.stderr b/test/train-sets/ref/sequencespan_data.nonldf.train.stderr
index a75bbfd5..2ea5e7e1 100644
--- a/test/train-sets/ref/sequencespan_data.nonldf.train.stderr
+++ b/test/train-sets/ref/sequencespan_data.nonldf.train.stderr
@@ -1,10 +1,10 @@
 final_regressor = models/sequencespan_data.model
+no rollout!
 Num weight bits = 18
 learning rate = 10
 initial_t = 1
 power_t = 0.5
 decay_learning_rate = 1
-no rollout!
 creating cache_file = train-sets/sequencespan_data.cache
 Reading datafile = train-sets/sequencespan_data
 num sources = 1
diff --git a/vowpalwabbit/cb_algs.cc b/vowpalwabbit/cb_algs.cc
index 2d8a2252..c676f547 100644
--- a/vowpalwabbit/cb_algs.cc
+++ b/vowpalwabbit/cb_algs.cc
@@ -496,7 +496,7 @@ namespace CB_ALGS
       *all.file_options << " --cb_type dr";
     }
 
-    if (!vm.count("csoaa"))
+    if (count(all.args.begin(), all.args.end(),"--csoaa") == 0)
       {
 	all.args.push_back("--csoaa");
 	stringstream ss;
@@ -524,7 +524,7 @@ namespace CB_ALGS
       }
     // preserve the increment of the base learner since we are
     // _adding_ to the number of problems rather than multiplying.
-    l->increment = all.l->increment; 
+    l->increment = base->increment; 
     
     l->set_init_driver(init_driver);
     l->set_finish(finish);
diff --git a/vowpalwabbit/cbify.cc b/vowpalwabbit/cbify.cc
index 776501f7..07e8fba1 100644
--- a/vowpalwabbit/cbify.cc
+++ b/vowpalwabbit/cbify.cc
@@ -390,7 +390,7 @@ namespace CBIFY {
     data.k = (uint32_t)vm["cbify"].as<size_t>();
     *all.file_options << " --cbify " << data.k;
 
-    if (!vm.count("cb"))
+    if (count(all.args.begin(), all.args.end(),"--cb") == 0)
       {
 	all.args.push_back("--cb");
 	stringstream ss;
diff --git a/vowpalwabbit/gd_mf.cc b/vowpalwabbit/gd_mf.cc
index 1a9a6335..4e91aeba 100644
--- a/vowpalwabbit/gd_mf.cc
+++ b/vowpalwabbit/gd_mf.cc
@@ -26,10 +26,12 @@ using namespace LEARNER;
 namespace GDMF {
   struct gdmf {
     vw* all;
+    uint32_t rank;
   };
 
-void mf_print_offset_features(vw& all, example& ec, size_t offset)
+void mf_print_offset_features(gdmf& d, example& ec, size_t offset)
 {
+  vw& all = *d.all;
   weight* weights = all.reg.weight_vector;
   size_t mask = all.reg.weight_mask;
   for (unsigned char* i = ec.indices.begin; i != ec.indices.end; i++) 
@@ -53,7 +55,7 @@ void mf_print_offset_features(vw& all, example& ec, size_t offset)
     if (ec.atomics[(int)(*i)[0]].size() > 0 && ec.atomics[(int)(*i)[1]].size() > 0)
       {
 	/* print out nsk^feature:hash:value:weight:nsk^feature^:hash:value:weight:prod_weights */
-	for (size_t k = 1; k <= all.rank; k++)
+	for (size_t k = 1; k <= d.rank; k++)
 	  {
 	    for (audit_data* f = ec.audit_features[(int)(*i)[0]].begin; f!= ec.audit_features[(int)(*i)[0]].end; f++)
 	      for (audit_data* f2 = ec.audit_features[(int)(*i)[1]].begin; f2!= ec.audit_features[(int)(*i)[1]].end; f2++)
@@ -62,11 +64,11 @@ void mf_print_offset_features(vw& all, example& ec, size_t offset)
 		       <<"(" << ((f->weight_index + offset +k) & mask)  << ")" << ':' << f->x;
 		  cout << ':' << weights[(f->weight_index + offset + k) & mask];
 		  
-		  cout << ':' << f2->space << k << '^' << f2->feature << ':' << ((f2->weight_index+k+all.rank)&mask) 
-		       <<"(" << ((f2->weight_index + offset +k+all.rank) & mask)  << ")" << ':' << f2->x;
-		  cout << ':' << weights[(f2->weight_index + offset + k+all.rank) & mask];
+		  cout << ':' << f2->space << k << '^' << f2->feature << ':' << ((f2->weight_index+k+d.rank)&mask) 
+		       <<"(" << ((f2->weight_index + offset +k+d.rank) & mask)  << ")" << ':' << f2->x;
+		  cout << ':' << weights[(f2->weight_index + offset + k+d.rank) & mask];
 		  
-		  cout << ':' <<  weights[(f->weight_index + offset + k) & mask] * weights[(f2->weight_index + offset + k + all.rank) & mask];
+		  cout << ':' <<  weights[(f->weight_index + offset + k) & mask] * weights[(f2->weight_index + offset + k + d.rank) & mask];
 		}
 	  }
       }
@@ -77,17 +79,25 @@ void mf_print_offset_features(vw& all, example& ec, size_t offset)
   cout << endl;
 }
 
-void mf_print_audit_features(vw& all, example& ec, size_t offset)
+void mf_print_audit_features(gdmf& d, example& ec, size_t offset)
 {
-  print_result(all.stdout_fileno,ec.pred.scalar,-1,ec.tag);
-  mf_print_offset_features(all, ec, offset);
+  print_result(d.all->stdout_fileno,ec.pred.scalar,-1,ec.tag);
+  mf_print_offset_features(d, ec, offset);
 }
 
-float mf_predict(vw& all, example& ec)
+float mf_predict(gdmf& d, example& ec)
 {
+  vw& all = *d.all;
   label_data& ld = ec.l.simple;
   float prediction = ld.initial;
 
+  for (vector<string>::iterator i = d.all->pairs.begin(); i != d.all->pairs.end();i++)
+    {
+      ec.num_features -= ec.atomics[(int)(*i)[0]].size() * ec.atomics[(int)(*i)[1]].size();
+      ec.num_features += ec.atomics[(int)(*i)[0]].size() * d.rank;
+      ec.num_features += ec.atomics[(int)(*i)[1]].size() * d.rank;
+    }
+
   // clear stored predictions
   ec.topic_predictions.erase();
 
@@ -107,18 +117,18 @@ float mf_predict(vw& all, example& ec)
     {
       if (ec.atomics[(int)(*i)[0]].size() > 0 && ec.atomics[(int)(*i)[1]].size() > 0)
 	{
-	  for (uint32_t k = 1; k <= all.rank; k++)
+	  for (uint32_t k = 1; k <= d.rank; k++)
 	    {
 	      // x_l * l^k
-	      // l^k is from index+1 to index+all.rank
+	      // l^k is from index+1 to index+d.rank
 	      //float x_dot_l = sd_offset_add(weights, mask, ec.atomics[(int)(*i)[0]].begin, ec.atomics[(int)(*i)[0]].end, k);
               float x_dot_l = 0.;
 	      GD::foreach_feature<float, GD::vec_add>(all.reg.weight_vector, all.reg.weight_mask, ec.atomics[(int)(*i)[0]].begin, ec.atomics[(int)(*i)[0]].end, x_dot_l, k);
 	      // x_r * r^k
-	      // r^k is from index+all.rank+1 to index+2*all.rank
-	      //float x_dot_r = sd_offset_add(weights, mask, ec.atomics[(int)(*i)[1]].begin, ec.atomics[(int)(*i)[1]].end, k+all.rank);
+	      // r^k is from index+d.rank+1 to index+2*d.rank
+	      //float x_dot_r = sd_offset_add(weights, mask, ec.atomics[(int)(*i)[1]].begin, ec.atomics[(int)(*i)[1]].end, k+d.rank);
               float x_dot_r = 0.;
-	      GD::foreach_feature<float,GD::vec_add>(all.reg.weight_vector, all.reg.weight_mask, ec.atomics[(int)(*i)[1]].begin, ec.atomics[(int)(*i)[1]].end, x_dot_r, k+all.rank);
+	      GD::foreach_feature<float,GD::vec_add>(all.reg.weight_vector, all.reg.weight_mask, ec.atomics[(int)(*i)[1]].begin, ec.atomics[(int)(*i)[1]].end, x_dot_r, k+d.rank);
 
 	      prediction += x_dot_l * x_dot_r;
 
@@ -146,7 +156,7 @@ float mf_predict(vw& all, example& ec)
     ec.loss = all.loss->getLoss(all.sd, ec.pred.scalar, ld.label) * ld.weight;
   
   if (all.audit)
-    mf_print_audit_features(all, ec, 0);
+    mf_print_audit_features(d, ec, 0);
   
   return ec.pred.scalar;
 }
@@ -158,55 +168,55 @@ void sd_offset_update(weight* weights, size_t mask, feature* begin, feature* end
     weights[(f->weight_index + offset) & mask] += update * f->x - regularization * weights[(f->weight_index + offset) & mask];
 }
 
-void mf_train(vw& all, example& ec)
-{
-      weight* weights = all.reg.weight_vector;
-      size_t mask = all.reg.weight_mask;
-      label_data& ld = ec.l.simple;
-
-      // use final prediction to get update size
+  void mf_train(gdmf& d, example& ec)
+  {
+    vw& all = *d.all;
+    weight* weights = all.reg.weight_vector;
+    size_t mask = all.reg.weight_mask;
+    label_data& ld = ec.l.simple;
+    
+    // use final prediction to get update size
       // update = eta_t*(y-y_hat) where eta_t = eta/(3*t^p) * importance weight
-      float eta_t = all.eta/pow(ec.example_t,all.power_t) / 3.f * ld.weight;
-      float update = all.loss->getUpdate(ec.pred.scalar, ld.label, eta_t, 1.); //ec.total_sum_feat_sq);
-
-      float regularization = eta_t * all.l2_lambda;
-
-      // linear update
-      for (unsigned char* i = ec.indices.begin; i != ec.indices.end; i++) 
-	sd_offset_update(weights, mask, ec.atomics[*i].begin, ec.atomics[*i].end, 0, update, regularization);
-      
-      // quadratic update
-      for (vector<string>::iterator i = all.pairs.begin(); i != all.pairs.end();i++) 
-	{
-	  if (ec.atomics[(int)(*i)[0]].size() > 0 && ec.atomics[(int)(*i)[1]].size() > 0)
-	    {
-
-	      // update l^k weights
-	      for (size_t k = 1; k <= all.rank; k++)
-		{
-		  // r^k \cdot x_r
-		  float r_dot_x = ec.topic_predictions[2*k];
-		  // l^k <- l^k + update * (r^k \cdot x_r) * x_l
-		  sd_offset_update(weights, mask, ec.atomics[(int)(*i)[0]].begin, ec.atomics[(int)(*i)[0]].end, k, update*r_dot_x, regularization);
-		}
-
-	      // update r^k weights
-	      for (size_t k = 1; k <= all.rank; k++)
-		{
-		  // l^k \cdot x_l
-		  float l_dot_x = ec.topic_predictions[2*k-1];
-		  // r^k <- r^k + update * (l^k \cdot x_l) * x_r
-		  sd_offset_update(weights, mask, ec.atomics[(int)(*i)[1]].begin, ec.atomics[(int)(*i)[1]].end, k+all.rank, update*l_dot_x, regularization);
-		}
-
-	    }
-	}
-  if (all.triples.begin() != all.triples.end()) {
-    cerr << "cannot use triples in matrix factorization" << endl;
-    throw exception();
-  }
-}  
-
+    float eta_t = all.eta/pow(ec.example_t,all.power_t) / 3.f * ld.weight;
+    float update = all.loss->getUpdate(ec.pred.scalar, ld.label, eta_t, 1.); //ec.total_sum_feat_sq);
+    
+    float regularization = eta_t * all.l2_lambda;
+    
+    // linear update
+    for (unsigned char* i = ec.indices.begin; i != ec.indices.end; i++) 
+      sd_offset_update(weights, mask, ec.atomics[*i].begin, ec.atomics[*i].end, 0, update, regularization);
+    
+    // quadratic update
+    for (vector<string>::iterator i = all.pairs.begin(); i != all.pairs.end();i++) 
+      {
+	if (ec.atomics[(int)(*i)[0]].size() > 0 && ec.atomics[(int)(*i)[1]].size() > 0)
+	  {
+	    
+	    // update l^k weights
+	    for (size_t k = 1; k <= d.rank; k++)
+	      {
+		// r^k \cdot x_r
+		float r_dot_x = ec.topic_predictions[2*k];
+		// l^k <- l^k + update * (r^k \cdot x_r) * x_l
+		sd_offset_update(weights, mask, ec.atomics[(int)(*i)[0]].begin, ec.atomics[(int)(*i)[0]].end, k, update*r_dot_x, regularization);
+	      }
+	    // update r^k weights
+	    for (size_t k = 1; k <= d.rank; k++)
+	      {
+		// l^k \cdot x_l
+		float l_dot_x = ec.topic_predictions[2*k-1];
+		// r^k <- r^k + update * (l^k \cdot x_l) * x_r
+		sd_offset_update(weights, mask, ec.atomics[(int)(*i)[1]].begin, ec.atomics[(int)(*i)[1]].end, k+d.rank, update*l_dot_x, regularization);
+	      }
+	    
+	  }
+      }
+    if (all.triples.begin() != all.triples.end()) {
+      cerr << "cannot use triples in matrix factorization" << endl;
+      throw exception();
+    }
+  }  
+  
   void save_load(gdmf& d, io_buf& model_file, bool read, bool text)
 {
   vw* all = d.all;
@@ -231,7 +241,7 @@ void mf_train(vw& all, example& ec)
       do 
 	{
 	  brw = 0;
-	  size_t K = all->rank*2+1;
+	  size_t K = d.rank*2+1;
 	  
 	  text_len = sprintf(buff, "%d ", i);
 	  brw += bin_text_read_write_fixed(model_file,(char *)&i, sizeof (i),
@@ -273,19 +283,16 @@ void mf_train(vw& all, example& ec)
   }
 
   void predict(gdmf& d, base_learner& base, example& ec)
-  {
-    vw* all = d.all;
- 
-    mf_predict(*all,ec);
+  { mf_predict(d,ec); 
   }
 
   void learn(gdmf& d, base_learner& base, example& ec)
   {
-    vw* all = d.all;
+    vw& all = *d.all;
  
-    predict(d, base, ec);
-    if (all->training && ec.l.simple.label != FLT_MAX)
-      mf_train(*all, ec);
+    mf_predict(d, ec);
+    if (all.training && ec.l.simple.label != FLT_MAX)
+      mf_train(d, ec);
   }
 
   base_learner* setup(vw& all, po::variables_map& vm)
@@ -294,16 +301,16 @@ void mf_train(vw& all, example& ec)
     opts.add_options()
       ("rank", po::value<uint32_t>(), "rank for matrix factorization.");
     vm = add_options(all, opts);
-    if(!vm.count("gdmf"))
+    if(!vm.count("rank"))
       return NULL;
-    else
-      all.rank = vm["gdmf"].as<uint32_t>();
 
     gdmf& data = calloc_or_die<gdmf>(); 
     data.all = &all;
+    data.rank = vm["rank"].as<uint32_t>();
 
+    *all.file_options << " --rank " << data.rank;
     // store linear + 2*rank weights per index, round up to power of two
-    float temp = ceilf(logf((float)(all.rank*2+1)) / logf (2.f));
+    float temp = ceilf(logf((float)(data.rank*2+1)) / logf (2.f));
     all.reg.stride_shift = (size_t) temp;
     all.random_weights = true;
     
diff --git a/vowpalwabbit/global_data.cc b/vowpalwabbit/global_data.cc
index bc869477..243e8657 100644
--- a/vowpalwabbit/global_data.cc
+++ b/vowpalwabbit/global_data.cc
@@ -271,7 +271,6 @@ vw::vw()
   power_t = 0.5;
   eta = 0.5; //default learning rate for normalized adaptive updates, this is switched to 10 by default for the other updates (see parse_args.cc)
   numpasses = 1;
-  rank = 0;
 
   final_prediction_sink.begin = final_prediction_sink.end=final_prediction_sink.end_array = NULL;
   raw_prediction = -1;
diff --git a/vowpalwabbit/global_data.h b/vowpalwabbit/global_data.h
index 6937c181..09ff643c 100644
--- a/vowpalwabbit/global_data.h
+++ b/vowpalwabbit/global_data.h
@@ -263,7 +263,6 @@ struct vw {
   std::string span_server;
 
   size_t length () { return ((size_t)1) << num_bits; };
-  uint32_t rank;
 
   v_array<LEARNER::base_learner* (*)(vw&, po::variables_map&)> reduction_stack;
 
diff --git a/vowpalwabbit/kernel_svm.cc b/vowpalwabbit/kernel_svm.cc
index a8773d64..272be528 100644
--- a/vowpalwabbit/kernel_svm.cc
+++ b/vowpalwabbit/kernel_svm.cc
@@ -795,8 +795,8 @@ namespace KSVM
     opts.add_options()
       ("ksvm", "kernel svm")
       ("reprocess", po::value<size_t>(), "number of reprocess steps for LASVM")
-      ("active", "do active learning")
-      ("active_c", po::value<double>(), "parameter for query prob")
+      //      ("active", "do active learning")
+      //("active_c", po::value<double>(), "parameter for query prob")
       ("pool_greedy", "use greedy selection on mini pools")      
       ("para_active", "do parallel active learning")
       ("pool_size", po::value<size_t>(), "size of pools for active learning")
diff --git a/vowpalwabbit/nn.cc b/vowpalwabbit/nn.cc
index a7079c95..44876e67 100644
--- a/vowpalwabbit/nn.cc
+++ b/vowpalwabbit/nn.cc
@@ -365,8 +365,10 @@ CONVERSE: // That's right, I'm using goto.  So sue me.
       n.xsubi = vm["random_seed"].as<size_t>();
 
     n.save_xsubi = n.xsubi;
-    n.increment = all.l->increment;//Indexing of output layer is odd.
-    learner<nn>& l = init_learner(&n,  setup_base(all,vm), predict_or_learn<true>, 
+    
+    base_learner* base = setup_base(all,vm);
+    n.increment = base->increment;//Indexing of output layer is odd.
+    learner<nn>& l = init_learner(&n, base, predict_or_learn<true>, 
 				  predict_or_learn<false>, n.k+1);
     l.set_finish(finish);
     l.set_finish_example(finish_example);
diff --git a/vowpalwabbit/parse_args.cc b/vowpalwabbit/parse_args.cc
index 0781830c..00b2a1d6 100644
--- a/vowpalwabbit/parse_args.cc
+++ b/vowpalwabbit/parse_args.cc
@@ -884,8 +884,6 @@ vw* parse_args(int argc, char *argv[])
       cerr << "power_t = " << all->power_t << endl;
       if (all->numpasses > 1)
 	cerr << "decay_learning_rate = " << all->eta_decay_rate << endl;
-      if (all->rank > 0)
-	cerr << "rank = " << all->rank << endl;
     }
 
   parse_output_preds(*all, vm);
diff --git a/vowpalwabbit/parse_regressor.cc b/vowpalwabbit/parse_regressor.cc
index def3a8de..6e7cf27e 100644
--- a/vowpalwabbit/parse_regressor.cc
+++ b/vowpalwabbit/parse_regressor.cc
@@ -164,11 +164,6 @@ void save_load_header(vw& all, io_buf& model_file, bool read, bool text)
 				"", read, 
 				"\n",1, text);
       
-      text_len = sprintf(buff, "rank:%d\n", (int)all.rank);
-      bin_text_read_write_fixed(model_file,(char*)&all.rank, sizeof(all.rank), 
-				"", read, 
-				buff,text_len, text);
-      
       text_len = sprintf(buff, "lda:%d\n", (int)all.lda);
       bin_text_read_write_fixed(model_file,(char*)&all.lda, sizeof(all.lda), 
 				"", read, 
diff --git a/vowpalwabbit/parser.cc b/vowpalwabbit/parser.cc
index 44055bab..65596402 100644
--- a/vowpalwabbit/parser.cc
+++ b/vowpalwabbit/parser.cc
@@ -840,37 +840,22 @@ void setup_example(vw& all, example* ae)
       ae->total_sum_feat_sq += ae->sum_feat_sq[*i];
     }
 
-  if (all.rank == 0) {
-    for (vector<string>::iterator i = all.pairs.begin(); i != all.pairs.end();i++)
-      {
-	ae->num_features 
-	  += ae->atomics[(int)(*i)[0]].size()
-	    *ae->atomics[(int)(*i)[1]].size();
-	ae->total_sum_feat_sq += ae->sum_feat_sq[(int)(*i)[0]]*ae->sum_feat_sq[(int)(*i)[1]];
-      }
-
-    for (vector<string>::iterator i = all.triples.begin(); i != all.triples.end();i++)
-      {
-	ae->num_features 
-	  += ae->atomics[(int)(*i)[0]].size()
-	    *ae->atomics[(int)(*i)[1]].size()
-	    *ae->atomics[(int)(*i)[2]].size();
-	ae->total_sum_feat_sq += ae->sum_feat_sq[(int)(*i)[0]] * ae->sum_feat_sq[(int)(*i)[1]] * ae->sum_feat_sq[(int)(*i)[2]];
-      }
-
-  } else {
-    for (vector<string>::iterator i = all.pairs.begin(); i != all.pairs.end();i++)
-      {
-	ae->num_features += ae->atomics[(int)(*i)[0]].size() * all.rank;
-	ae->num_features += ae->atomics[(int)(*i)[1]].size() * all.rank;
-      }
-    for (vector<string>::iterator i = all.triples.begin(); i != all.triples.end();i++)
-      {
-	ae->num_features += ae->atomics[(int)(*i)[0]].size() * all.rank;
-	ae->num_features += ae->atomics[(int)(*i)[1]].size() * all.rank;
-	ae->num_features += ae->atomics[(int)(*i)[2]].size() * all.rank;
-      }
-  }
+  for (vector<string>::iterator i = all.pairs.begin(); i != all.pairs.end();i++)
+    {
+      ae->num_features 
+	+= ae->atomics[(int)(*i)[0]].size()
+	*ae->atomics[(int)(*i)[1]].size();
+      ae->total_sum_feat_sq += ae->sum_feat_sq[(int)(*i)[0]]*ae->sum_feat_sq[(int)(*i)[1]];
+    }
+  
+  for (vector<string>::iterator i = all.triples.begin(); i != all.triples.end();i++)
+    {
+      ae->num_features 
+	+= ae->atomics[(int)(*i)[0]].size()
+	*ae->atomics[(int)(*i)[1]].size()
+	*ae->atomics[(int)(*i)[2]].size();
+      ae->total_sum_feat_sq += ae->sum_feat_sq[(int)(*i)[0]] * ae->sum_feat_sq[(int)(*i)[1]] * ae->sum_feat_sq[(int)(*i)[2]];
+    }
 }
 }
 
diff --git a/vowpalwabbit/search.cc b/vowpalwabbit/search.cc
index 5e782ea1..5bc7c797 100644
--- a/vowpalwabbit/search.cc
+++ b/vowpalwabbit/search.cc
@@ -1956,7 +1956,10 @@ namespace Search {
     }
     all.p->emptylines_separate_examples = true;
 
-    if (!vm.count("csoaa") && !vm.count("csoaa_ldf") && !vm.count("wap_ldf") && !vm.count("cb"))
+    if (count(all.args.begin(), all.args.end(),"--csoaa") == 0 
+	&& count(all.args.begin(), all.args.end(),"--csoaa_ldf") == 0 
+	&& count(all.args.begin(), all.args.end(),"--wap_ldf") == 0 
+	&&  count(all.args.begin(), all.args.end(),"--cb") == 0)
       {
 	all.args.push_back("--csoaa");
 	stringstream ss;