refactor vest mapper input; --optimize_feature (s) should now limit non-oracle directions

git-svn-id: https://ws10smt.googlecode.com/svn/trunk@273 ec762483-ff6d-05da-a07a-a48fb63a330f

1 files changed, 163 insertions, 107 deletions

diff --git a/vest/mr_vest_generate_mapper_input.cc b/vest/mr_vest_generate_mapper_input.cc
index c0f80d0c..e9a5650b 100644
--- a/vest/mr_vest_generate_mapper_input.cc
+++ b/vest/mr_vest_generate_mapper_input.cc
@@ -13,15 +13,149 @@
 #include "hg_io.h"
 #include "scorer.h"
 #include "oracle_bleu.h"
+#include "ff_bleu.h"
+
+boost::shared_ptr<FFRegistry> global_ff_registry;
+namespace {
+struct init_ff {
+  init_ff() {
+    global_ff_registry.reset(new FFRegistry);
+    global_ff_registry->Register(new FFFactory<BLEUModel>);
+  }
+};
+init_ff reg;
+}
 
 using namespace std;
 namespace po = boost::program_options;
 
 typedef SparseVector<double> Dir;
+typedef Dir Point;
 
-MT19937 rng;
+
+void compress_similar(vector<Dir> &dirs,double min_dist,ostream *log=&cerr,bool avg=true) {
+  if (min_dist<=0) return;
+  double max_s=1.-min_dist;
+  unsigned N=dirs.size();
+  for (int i=0;i<N;++i) {
+    for (int j=i+1;j<N;++j) {
+      double s=dirs[i].tanimoto_coef(dirs[j]);
+      if (s>max_s) {
+        if (log) *log << "Collapsing similar directions (T="<<s<<" > "<<max_s<<").  dirs["<<i<<"]="<<dirs[i]<<" dirs["<<j<<"]";
+        if (avg) {
+          dirs[i]+=dirs[j];
+          dirs[i]/=2.;
+          if (log) *log<<" averaged="<<dirs[i];
+        }
+        if (log) *log<<endl;
+        swap(dirs[j],dirs[--N]);
+      }
+    }
+  }
+  dirs.resize(N);
+}
 
 struct oracle_directions {
+  MT19937 rng;
+  OracleBleu oracle;
+  vector<Dir> directions;
+
+  bool start_random;
+  bool include_primary;
+  bool fear_to_hope;
+  unsigned n_random;
+  void AddPrimaryAndRandomDirections() {
+    LineOptimizer::CreateOptimizationDirections(
+      fids,n_random,&rng,&directions,include_primary);
+  }
+
+  void Print() {
+    for (int i = 0; i < dev_set_size; ++i)
+      for (int j = 0; j < directions.size(); ++j)
+        cout << forest_file(i) <<" " << i << ' ' << origin << ' ' << directions[j] << endl;
+  }
+
+  void AddOptions(po::options_description *opts) {
+    oracle.AddOptions(opts);
+  }
+
+  void InitCommandLine(int argc, char *argv[], po::variables_map *conf) {
+    po::options_description opts("Configuration options");
+    OracleBleu::AddOptions(&opts);
+    opts.add_options()
+      ("dev_set_size,s",po::value<unsigned>(&dev_set_size),"[REQD] Development set size (# of parallel sentences)")
+      ("forest_repository,r",po::value<string>(),"[REQD] Path to forest repository")
+      ("weights,w",po::value<string>(),"[REQD] Current feature weights file")
+      ("optimize_feature,o",po::value<vector<string> >(), "Feature to optimize (if none specified, all weights listed in the weights file will be optimized)")
+      ("random_directions,d",po::value<unsigned int>()->default_value(20),"Number of random directions to run the line optimizer in")
+      ("no_primary,n","don't use the primary (orthogonal each feature alone) directions")
+      ("oracle_directions,O",po::value<unsigned>()->default_value(0),"read the forests and choose this many directions based on heading toward a hope max (bleu+modelscore) translation.")
+      ("oracle_start_random",po::bool_switch(&start_random),"sample random subsets of dev set for ALL oracle directions, not just those after a sequential run through it")
+      ("oracle_batch,b",po::value<unsigned>()->default_value(10),"to produce each oracle direction, sum the 'gradient' over this many sentences")
+      ("max_similarity,m",po::value<double>()->default_value(0),"remove directions that are too similar (Tanimoto coeff. less than (1-this)).  0 means don't filter, 1 means only 1 direction allowed?")
+      ("fear_to_hope,f","for each of the oracle_directions, also include a direction from fear to hope (as well as origin to hope)")
+      ("help,h", "Help");
+    po::options_description dcmdline_options;
+    dcmdline_options.add(opts);
+    po::store(parse_command_line(argc, argv, dcmdline_options), *conf);
+    bool flag = false;
+    if (conf->count("dev_set_size") == 0) {
+      cerr << "Please specify the size of the development set using -d N\n";
+      flag = true;
+    }
+    if (conf->count("weights") == 0) {
+      cerr << "Please specify the starting-point weights using -w <weightfile.txt>\n";
+      flag = true;
+    }
+    if (conf->count("forest_repository") == 0) {
+      cerr << "Please specify the forest repository location using -r <DIR>\n";
+      flag = true;
+    }
+    if (flag || conf->count("help")) {
+      cerr << dcmdline_options << endl;
+      exit(1);
+    }
+  }
+
+  int main(int argc, char *argv[]) {
+    po::variables_map conf;
+    InitCommandLine(argc,argv,&conf);
+    UseConf(conf);
+    Run();
+    return 0;
+  }
+
+  void Run() {
+    AddPrimaryAndRandomDirections();
+    AddOracleDirections();
+    compress_similar(directions,max_similarity);
+    Print();
+  }
+
+
+  Point origin; // old weights that gave model 1best.
+  vector<string> optimize_features;
+  void UseConf(po::variables_map const& conf) {
+    oracle.UseConf(conf);
+
+    include_primary=!conf.count("no_primary");
+    if (conf.count("optimize_feature") > 0)
+      optimize_features=conf["optimize_feature"].as<vector<string> >();
+    fear_to_hope=conf.count("fear_to_hope");
+    n_random=conf["random_directions"].as<unsigned int>();
+    forest_repository=conf["forest_repository"].as<string>();
+//    dev_set_size=conf["dev_set_size"].as<unsigned int>();
+    n_oracle=conf["oracle_directions"].as<unsigned>();
+    oracle_batch=conf["oracle_batch"].as<unsigned>();
+    max_similarity=conf["max_similarity"].as<double>();
+    weights_file=conf["weights"].as<string>();
+
+    Init();
+  }
+
+  string weights_file;
+  double max_similarity;
+  unsigned n_oracle, oracle_batch;
   string forest_repository;
   unsigned dev_set_size;
   vector<Dir> dirs; //best_to_hope_dirs
@@ -32,14 +166,28 @@ struct oracle_directions {
     return o.str();
   }
 
-  void set_dev_set_size(int i) {
-    dev_set_size=i;
-    dirs.resize(dev_set_size);
+  oracle_directions() { }
+
+  void Init() {
+    start_random=false;
+    assert(DirectoryExists(forest_repository));
+    vector<string> features;
+    weights.InitFromFile(weights_file, &features);
+    if (optimize_features.size())
+      features=optimize_features;
+    weights.InitSparseVector(&origin);
+    fids.clear();
+    AddFeatureIds(features);
   }
 
-  oracle_directions(string forest_repository="",unsigned dev_set_sz=0,vector<int> const& fids=vector<int>()): forest_repository(forest_repository),fids(fids) {
-    set_dev_set_size(dev_set_sz);
-  }
+  Weights weights;
+  void AddFeatureIds(vector<string> const& features) {
+    int i = fids.size();
+    fids.resize(fids.size()+features.size());
+    for (; i < features.size(); ++i)
+      fids[i] = FD::Convert(features[i]);
+ }
+
 
   Dir const& operator[](unsigned i) {
     Dir &dir=dirs[i];
@@ -52,112 +200,20 @@ struct oracle_directions {
     return dir;
   }
   // if start_random is true, immediately sample w/ replacement from src sentences; otherwise, consume them sequentially until exhausted, then random.  oracle vectors are summed
-  void add_directions(vector<Dir> &dirs,unsigned n,unsigned batchsz=20,bool start_random=false) {
+  void AddOracleDirections() {
     MT19937::IntRNG rsg=rng.inclusive(0,dev_set_size-1);
     unsigned b=0;
-    for(unsigned i=0;i<n;++i) {
-      dirs.push_back(Dir());
-      Dir &d=dirs.back();
-      for (unsigned j=0;j<batchsz;++j,++b)
+    for(unsigned i=0;i<n_oracle;++i) {
+      directions.push_back(Dir());
+      Dir &d=directions.back();
+      for (unsigned j=0;j<oracle_batch;++j,++b)
         d+=(*this)[(start_random || b>=dev_set_size)?rsg():b];
-      d/=(double)batchsz;
+      d/=(double)oracle_batch;
     }
   }
-
 };
 
-void compress_similar(vector<Dir> &dirs,double min_dist,ostream *log=&cerr,bool avg=true) {
-  if (min_dist<=0) return;
-  double max_s=1.-min_dist;
-  unsigned N=dirs.size();
-  for (int i=0;i<N;++i) {
-    for (int j=i+1;j<N;++j) {
-      double s=dirs[i].tanimoto_coef(dirs[j]);
-      if (s>max_s) {
-        if (log) *log << "Collapsing similar directions (T="<<s<<" > "<<max_s<<").  dirs["<<i<<"]="<<dirs[i]<<" dirs["<<j<<"]";
-        if (avg) {
-          dirs[i]+=dirs[j];
-          dirs[i]/=2.;
-          if (log) *log<<" averaged="<<dirs[i];
-        }
-        if (log) *log<<endl;
-        swap(dirs[j],dirs[--N]);
-      }
-    }
-  }
-  dirs.resize(N);
-}
-
-
-
-void InitCommandLine(int argc, char** argv, po::variables_map* conf) {
-  po::options_description opts("Configuration options");
-  OracleBleu::AddOptions(&opts);
-  opts.add_options()
-        ("dev_set_size,s",po::value<unsigned int>(),"[REQD] Development set size (# of parallel sentences)")
-        ("forest_repository,r",po::value<string>(),"[REQD] Path to forest repository")
-        ("weights,w",po::value<string>(),"[REQD] Current feature weights file")
-        ("optimize_feature,o",po::value<vector<string> >(), "Feature to optimize (if none specified, all weights listed in the weights file will be optimized)")
-        ("random_directions,d",po::value<unsigned int>()->default_value(20),"Number of random directions to run the line optimizer in")
-    ("no_primary,n","don't use the primary (orthogonal each feature alone) directions")
-    ("oracle_directions,O",po::value<unsigned>()->default_value(0),"read the forests and choose this many directions based on heading toward a hope max (bleu+modelscore) translation.")
-    ("oracle_batch,b",po::value<unsigned>()->default_value(10),"to produce each oracle direction, sum the 'gradient' over this many sentences")
-    ("max_similarity,m",po::value<double>()->default_value(0),"remove directions that are too similar (Tanimoto coeff. less than (1-this)).  0 means don't filter, 1 means only 1 direction allowed?")
-        ("help,h", "Help");
-  po::options_description dcmdline_options;
-  dcmdline_options.add(opts);
-  po::store(parse_command_line(argc, argv, dcmdline_options), *conf);
-  bool flag = false;
-  if (conf->count("dev_set_size") == 0) {
-    cerr << "Please specify the size of the development set using -d N\n";
-    flag = true;
-  }
-  if (conf->count("weights") == 0) {
-    cerr << "Please specify the starting-point weights using -w <weightfile.txt>\n";
-    flag = true;
-  }
-  if (conf->count("forest_repository") == 0) {
-    cerr << "Please specify the forest repository location using -r <DIR>\n";
-    flag = true;
-  }
-  if (flag || conf->count("help")) {
-    cerr << dcmdline_options << endl;
-    exit(1);
-  }
-}
-
 int main(int argc, char** argv) {
-  po::variables_map conf;
-  InitCommandLine(argc, argv, &conf);
-  Weights weights;
-  vector<string> features;
-  weights.InitFromFile(conf["weights"].as<string>(), &features);
-  vector<int> fids(features.size());
-  for (int i = 0; i < features.size(); ++i)
-    fids[i] = FD::Convert(features[i]);
-
-  oracle_directions od(conf["forest_repository"].as<string>()
-                       , conf["dev_set_size"].as<unsigned int>()
-                       , fids
-    );
-;
-  assert(DirectoryExists(od.forest_repository));
-  SparseVector<double> origin;
-  weights.InitSparseVector(&origin);
-  if (conf.count("optimize_feature") > 0)
-    features=conf["optimize_feature"].as<vector<string> >();
-  vector<SparseVector<double> > axes;
-  LineOptimizer::CreateOptimizationDirections(
-     fids,
-     conf["random_directions"].as<unsigned int>(),
-     &rng,
-     &axes,
-     !conf.count("no_primary")
-    );
-  od.add_directions(axes,conf["oracle_directions"].as<unsigned>(),conf["oracle_batch"].as<unsigned>());
-  compress_similar(axes,conf["max_similarity"].as<double>());
-  for (int i = 0; i < od.dev_set_size; ++i)
-    for (int j = 0; j < axes.size(); ++j)
-      cout << od.forest_file(i) <<" " << i << ' ' << origin << ' ' << axes[j] << endl;
-  return 0;
+  oracle_directions od;
+  return od.main(argc,argv);
 }