8. Classifiers

Contents

1 Introduction
2 Training data
3 Training

Introduction

TIPL includes a small machine learning library that provides several useful classifiers, including Naive Bayes, logistic regression, Adaboost, K-nearest neighbor.

Naive Bayes

// declare a Naive Bayes classifier

naive_bayes<double,unsigned char> NB;

Logistic regression

// declare a logistic regression classifier

logistic_regression<double,unsigned char> LR(0.1,0.0001);

Adaboost

// Boosting using decision stump

ada_boost<decision_stump<double,unsigned char> > boosted_decision1((size_t)10);

K-nearest neighbor

nearest_neighbor<double,unsigned char,weight_function_average> NN(1);

Parzen density estimator

parzen_estimator<double,unsigned char,weight_function_average> PE(1.0);

Decision tree

decision_tree<double,unsigned char> DE(100,0.9);

Training data

// prepare learning data
   // feature type: double
   // classification type: unsigned char
   training_data<double,unsigned char> data;

   // 1000 training data
   data.features.resize(1000);
   data.classification.resize(1000);
   unsigned int
attribute_dimension = 3;
   for(int index = 0;index < 1000;++index)

   {
   data.features[index].resize(attribute_dimension);

// input the features
   data.features[index][0] = 1.0;
   data.features[index][1] = 1.0;
   data.features[index][2] = 1.0;
       data.classification[index] = 0;
   }

Training

// train the Nearest neighbor classifier using all dataset
   NN.learn(data.features.begin(),

   data.features.end(),
   attribute_dimension
,
   data.classification.begin());

Comments

Contact