dlib C++ Library - sequence_segmenter.cpp

// Copyright (C) 2013 Davis E. King (davis@dlib.net)
// License: Boost Software License See LICENSE.txt for the full license.
#include <sstream>
#include "tester.h"
#include <dlib/svm_threaded.h>
#include <dlib/rand.h>
namespace 
{
 using namespace test;
 using namespace dlib;
 using namespace std;
 logger dlog("test.sequence_segmenter");
// ----------------------------------------------------------------------------------------
 dlib::rand rnd;
 template <bool use_BIO_model_, bool use_high_order_features_, bool allow_negative_weights_>
 class unigram_extractor
 {
 public:
 const static bool use_BIO_model = use_BIO_model_;
 const static bool use_high_order_features = use_high_order_features_;
 const static bool allow_negative_weights = allow_negative_weights_;
 typedef std::vector<unsigned long> sequence_type; 
 std::map<unsigned long, matrix<double,0,1> > feats;
 unigram_extractor()
 {
 matrix<double,0,1> v1, v2, v3;
 v1 = randm(num_features(), 1, rnd);
 v2 = randm(num_features(), 1, rnd);
 v3 = randm(num_features(), 1, rnd);
 v1(0) = 1;
 v2(1) = 1;
 v3(2) = 1;
 v1(3) = -1;
 v2(4) = -1;
 v3(5) = -1;
 for (unsigned long i = 0; i < num_features(); ++i)
 {
 if ( i < 3)
 feats[i] = v1;
 else if (i < 6)
 feats[i] = v2;
 else
 feats[i] = v3;
 }
 }
 unsigned long num_features() const { return 10; }
 unsigned long window_size() const { return 3; }
 template <typename feature_setter>
 void get_features (
 feature_setter& set_feature,
 const sequence_type& x,
 unsigned long position
 ) const
 {
 const matrix<double,0,1>& m = feats.find(x[position])->second;
 for (unsigned long i = 0; i < num_features(); ++i)
 {
 set_feature(i, m(i));
 }
 }
 };
 template <bool use_BIO_model_, bool use_high_order_features_, bool neg>
 void serialize(const unigram_extractor<use_BIO_model_,use_high_order_features_,neg>& item , std::ostream& out )
 {
 serialize(item.feats, out);
 }
 template <bool use_BIO_model_, bool use_high_order_features_, bool neg>
 void deserialize(unigram_extractor<use_BIO_model_,use_high_order_features_,neg>& item, std::istream& in)
 {
 deserialize(item.feats, in);
 }
// ----------------------------------------------------------------------------------------
 void make_dataset (
 std::vector<std::vector<unsigned long> >& samples,
 std::vector<std::vector<unsigned long> >& labels,
 unsigned long dataset_size
 )
 {
 samples.clear();
 labels.clear();
 samples.resize(dataset_size);
 labels.resize(dataset_size);
 unigram_extractor<true,true,true> fe;
 dlib::rand rnd;
 for (unsigned long iter = 0; iter < dataset_size; ++iter)
 {
 samples[iter].resize(10);
 labels[iter].resize(10);
 for (unsigned long i = 0; i < samples[iter].size(); ++i)
 {
 samples[iter][i] = rnd.get_random_32bit_number()%fe.num_features();
 if (samples[iter][i] < 3)
 {
 labels[iter][i] = impl_ss::BEGIN;
 }
 else if (samples[iter][i] < 6)
 {
 labels[iter][i] = impl_ss::INSIDE;
 }
 else
 {
 labels[iter][i] = impl_ss::OUTSIDE;
 }
 if (i != 0)
 {
 // do rejection sampling to avoid impossible labels
 if (labels[iter][i] == impl_ss::INSIDE &&
 labels[iter][i-1] == impl_ss::OUTSIDE)
 {
 --i;
 }
 }
 }
 }
 }
// ----------------------------------------------------------------------------------------
 void make_dataset2 (
 std::vector<std::vector<unsigned long> >& samples,
 std::vector<std::vector<std::pair<unsigned long, unsigned long> > >& segments,
 unsigned long dataset_size
 )
 {
 segments.clear();
 std::vector<std::vector<unsigned long> > labels;
 make_dataset(samples, labels, dataset_size);
 segments.resize(samples.size());
 // Convert from BIO tagging to the explicit segments representation.
 for (unsigned long k = 0; k < labels.size(); ++k)
 {
 for (unsigned long i = 0; i < labels[k].size(); ++i)
 {
 if (labels[k][i] == impl_ss::BEGIN)
 {
 const unsigned long begin = i;
 ++i;
 while (i < labels[k].size() && labels[k][i] == impl_ss::INSIDE)
 ++i;
 segments[k].push_back(std::make_pair(begin, i));
 --i;
 }
 }
 }
 }
// ----------------------------------------------------------------------------------------
 template <bool use_BIO_model, bool use_high_order_features, bool allow_negative_weights>
 void do_test()
 {
 dlog << LINFO << "use_BIO_model: "<< use_BIO_model;
 dlog << LINFO << "use_high_order_features: "<< use_high_order_features;
 dlog << LINFO << "allow_negative_weights: "<< allow_negative_weights;
 std::vector<std::vector<unsigned long> > samples;
 std::vector<std::vector<std::pair<unsigned long,unsigned long> > > segments;
 make_dataset2( samples, segments, 100);
 print_spinner();
 typedef unigram_extractor<use_BIO_model,use_high_order_features,allow_negative_weights> fe_type;
 fe_type fe_temp;
 fe_type fe_temp2;
 structural_sequence_segmentation_trainer<fe_type> trainer(fe_temp2);
 trainer.set_c(5);
 trainer.set_num_threads(1);
 sequence_segmenter<fe_type> labeler = trainer.train(samples, segments);
 print_spinner();
 const std::vector<std::pair<unsigned long, unsigned long> > predicted_labels = labeler(samples[1]);
 const std::vector<std::pair<unsigned long, unsigned long> > true_labels = segments[1];
 /*
 for (unsigned long i = 0; i < predicted_labels.size(); ++i)
 cout << "["<<predicted_labels[i].first<<","<<predicted_labels[i].second<<") ";
 cout << endl;
 for (unsigned long i = 0; i < true_labels.size(); ++i)
 cout << "["<<true_labels[i].first<<","<<true_labels[i].second<<") ";
 cout << endl;
 */
 DLIB_TEST(predicted_labels.size() > 0);
 DLIB_TEST(predicted_labels.size() == true_labels.size());
 for (unsigned long i = 0; i < predicted_labels.size(); ++i)
 {
 DLIB_TEST(predicted_labels[i].first == true_labels[i].first);
 DLIB_TEST(predicted_labels[i].second == true_labels[i].second);
 }
 matrix<double> res;
 res = cross_validate_sequence_segmenter(trainer, samples, segments, 3);
 dlog << LINFO << "cv res: "<< res;
 DLIB_TEST(min(res) > 0.98);
 make_dataset2( samples, segments, 100);
 res = test_sequence_segmenter(labeler, samples, segments);
 dlog << LINFO << "test res: "<< res;
 DLIB_TEST(min(res) > 0.98);
 print_spinner();
 ostringstream sout;
 serialize(labeler, sout);
 istringstream sin(sout.str());
 sequence_segmenter<fe_type> labeler2;
 deserialize(labeler2, sin);
 res = test_sequence_segmenter(labeler2, samples, segments);
 dlog << LINFO << "test res2: "<< res;
 DLIB_TEST(min(res) > 0.98);
 long N;
 if (use_BIO_model)
 N = 3*3+3;
 else
 N = 5*5+5;
 const double min_normal_weight = min(colm(labeler2.get_weights(), 0, labeler2.get_weights().size()-N));
 const double min_trans_weight = min(labeler2.get_weights());
 dlog << LINFO << "min_normal_weight: " << min_normal_weight;
 dlog << LINFO << "min_trans_weight: " << min_trans_weight;
 if (allow_negative_weights)
 {
 DLIB_TEST(min_normal_weight < 0);
 DLIB_TEST(min_trans_weight < 0);
 }
 else
 {
 DLIB_TEST(min_normal_weight == 0);
 DLIB_TEST(min_trans_weight < 0);
 }
 }
// ----------------------------------------------------------------------------------------
 class unit_test_sequence_segmenter : public tester
 {
 public:
 unit_test_sequence_segmenter (
 ) :
 tester ("test_sequence_segmenter",
 "Runs tests on the sequence segmenting code.")
 {}
 void perform_test (
 )
 {
 do_test<true,true,false>();
 do_test<true,false,false>();
 do_test<false,true,false>();
 do_test<false,false,false>();
 do_test<true,true,true>();
 do_test<true,false,true>();
 do_test<false,true,true>();
 do_test<false,false,true>();
 }
 } a;
}