Open-Harbor/osrm-backend
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 3 Wiki Activity

several copy edits to brush up the code

Browse Source 
- fix copyright header
- rename probabilityDensityFunction -> density_function
- use double-precision fp literal to indicate intent
- remove redundant enum class start value
- replace C-style comments with C++ style
- make functions const
This commit is contained in:
Dennis Luxen 2015-03-03 17:41:12 +01:00
parent 133e382aae
commit 31cae8f05f
1 changed files with 31 additions and 28 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

59
algorithms/bayes_classifier.hpp
View File

@ -1,6 +1,6 @@
/* /*
Copyright (c) 2015, Project OSRM, Dennis Luxen, others Copyright (c) 2015, Project OSRM contributors
All rights reserved. All rights reserved.
Redistribution and use in source and binary forms, with or without modification, Redistribution and use in source and binary forms, with or without modification,
@ -28,72 +28,75 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#ifndef BAYES_CLASSIFIER_HPP #ifndef BAYES_CLASSIFIER_HPP
#define BAYES_CLASSIFIER_HPP #define BAYES_CLASSIFIER_HPP
#include <vector>
#include <cmath> #include <cmath>
#include <vector>
struct NormalDistribution struct NormalDistribution
{ {
NormalDistribution(const double mean, const double standard_deviation) NormalDistribution(const double mean, const double standard_deviation)
: mean(mean) : mean(mean), standard_deviation(standard_deviation)
, standard_deviation(standard_deviation)
{ {
} }
// FIXME implement log-probability version since its faster // FIXME implement log-probability version since its faster
double probabilityDensityFunction(const double val) double density_function(const double val) const
{ {
const double x = val - mean; const double x = val - mean;
return 1.0 / (std::sqrt(2*M_PI) * standard_deviation) * std::exp(-x*x / (standard_deviation * standard_deviation)); return 1.0 / (std::sqrt(2. * M_PI) * standard_deviation) *
std::exp(-x * x / (standard_deviation * standard_deviation));
} }
double mean; double mean;
double standard_deviation; double standard_deviation;
}; };
struct LaplaceDistribution struct LaplaceDistribution
{ {
LaplaceDistribution(const double location, const double scale) LaplaceDistribution(const double location, const double scale)
: location(location) : location(location), scale(scale)
, scale(scale)
{ {
} }
// FIXME implement log-probability version since its faster // FIXME implement log-probability version since its faster
double probabilityDensityFunction(const double val) const double density_function(const double val) const
{ {
const double x = std::abs(val - location); const double x = std::abs(val - location);
return 1.0 / (2*scale) * std::exp(-x / scale); return 1.0 / (2. * scale) * std::exp(-x / scale);
} }
double location; double location;
double scale; double scale;
}; };
template<typename PositiveDistributionT, typename NegativeDistributionT, typename ValueT> template <typename PositiveDistributionT, typename NegativeDistributionT, typename ValueT>
class BayesClassifier class BayesClassifier
{ {
public: public:
enum class ClassLabel : unsigned {NEGATIVE = 0, POSITIVE}; enum class ClassLabel : unsigned
{
NEGATIVE,
POSITIVE
};
using ClassificationT = std::pair<ClassLabel, double>; using ClassificationT = std::pair<ClassLabel, double>;
BayesClassifier(const PositiveDistributionT& positive_distribution, BayesClassifier(const PositiveDistributionT &positive_distribution,
const NegativeDistributionT& negative_distribution, const NegativeDistributionT &negative_distribution,
const double positive_apriori_probability) const double positive_apriori_probability)
: positive_distribution(positive_distribution) : positive_distribution(positive_distribution),
, negative_distribution(negative_distribution) negative_distribution(negative_distribution),
, positive_apriori_probability(positive_apriori_probability) positive_apriori_probability(positive_apriori_probability),
, negative_apriori_probability(1 - positive_apriori_probability) negative_apriori_probability(1. - positive_apriori_probability)
{ {
} }
/* // Returns label and the probability of the label.
* Returns label and the probability of the label. ClassificationT classify(const ValueT &v) const
*/
ClassificationT classify(const ValueT& v) const
{ {
const double positive_postpriori = positive_apriori_probability * positive_distribution.probabilityDensityFunction(v); const double positive_postpriori =
const double negative_postpriori = negative_apriori_probability * negative_distribution.probabilityDensityFunction(v); positive_apriori_probability * positive_distribution.density_function(v);
const double negative_postpriori =
negative_apriori_probability * negative_distribution.density_function(v);
const double norm = positive_postpriori + negative_postpriori; const double norm = positive_postpriori + negative_postpriori;
if (positive_postpriori > negative_postpriori) if (positive_postpriori > negative_postpriori)
@ -104,11 +107,11 @@ public:
return std::make_pair(ClassLabel::NEGATIVE, negative_postpriori / norm); return std::make_pair(ClassLabel::NEGATIVE, negative_postpriori / norm);
} }
private: private:
PositiveDistributionT positive_distribution; PositiveDistributionT positive_distribution;
NegativeDistributionT negative_distribution; NegativeDistributionT negative_distribution;
double positive_apriori_probability; double positive_apriori_probability;
double negative_apriori_probability; double negative_apriori_probability;
}; };
#endif /* BAYES_CLASSIFIER_HPP */ #endif // BAYES_CLASSIFIER_HPP