Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Comments added

Browse Source
This commit is contained in:
kalenhaha 2014-02-13 13:04:55 +08:00
parent 06ce8c9f3a
commit f22139c659
7 changed files with 315 additions and 153 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
booster/xgboost_data.h
View File

@ -9,6 +9,7 @@
#include <vector>
#include "../utils/xgboost_utils.h"
#include "../utils/xgboost_stream.h"
namespace xgboost{
namespace booster{

146
regression/xgboost_reg.h
View File

@ -16,6 +16,11 @@ namespace xgboost{
/*! \brief class for gradient boosted regression */
class RegBoostLearner{
public:
RegBoostLearner(bool silent = false){
this->silent = silent;
}
/*!
* \brief a regression booter associated with training and evaluating data
* \param train pointer to the training data
@ -23,13 +28,34 @@ namespace xgboost{
* \param evname name of evaluation data, used print statistics
*/
RegBoostLearner( const DMatrix *train,
std::vector<const DMatrix *> evals,
std::vector<std::string> evname, bool silent = false ){
this->silent = silent;
SetData(train,evals,evname);
}
/*!
* \brief associate regression booster with training and evaluating data
* \param train pointer to the training data
* \param evals array of evaluating data
* \param evname name of evaluation data, used print statistics
*/
inline void SetData(const DMatrix *train,
std::vector<const DMatrix *> evals,
std::vector<std::string> evname){
this->train_ = train;
this->evals_ = evals;
this->evname_ = evname;
//TODO: assign buffer index
//assign buffer index
int buffer_size = (*train).size();
for(int i = 0; i < evals.size(); i++){
buffer_size += (*evals[i]).size();
}
char str[25];
itoa(buffer_size,str,10);
base_model.SetParam("num_pbuffer",str);
}
/*!
* \brief set parameters from outside
* \param name name of the parameter
@ -47,6 +73,14 @@ namespace xgboost{
base_model.InitTrainer();
mparam.AdjustBase();
}
/*!
* \brief initialize the current data storage for model, if the model is used first time, call this function
*/
inline void InitModel( void ){
base_model.InitModel();
}
/*!
* \brief load model from stream
* \param fi input stream
@ -63,23 +97,66 @@ namespace xgboost{
fo.Write( &mparam, sizeof(ModelParam) );
base_model.SaveModel( fo );
}
/*!
* \brief update the model for one iteration
* \param iteration the number of updating iteration
*/
inline void UpdateOneIter( void ){
//TODO
inline void UpdateOneIter( int iteration ){
std::vector<float> grad,hess,preds;
std::vector<unsigned> root_index;
booster::FMatrixS::Image train_image((*train_).data);
Predict(preds,*train_,0);
Gradient(preds,(*train_).labels,grad,hess);
base_model.DoBoost(grad,hess,train_image,root_index);
int buffer_index_offset = (*train_).size();
float loss = 0.0;
for(int i = 0; i < evals_.size();i++){
Predict(preds, *evals_[i], buffer_index_offset);
loss = mparam.Loss(preds,(*evals_[i]).labels);
if(!silent){
printf("The loss of %s data set in %d the \
iteration is %f",evname_[i].c_str(),&iteration,&loss);
}
/*! \brief predict the results, given data */
inline void Predict( std::vector<float> &preds, const DMatrix &data ){
//TODO
buffer_index_offset += (*evals_[i]).size();
}
}
/*! \brief get the transformed predictions, given data */
inline void Predict( std::vector<float> &preds, const DMatrix &data,int buffer_index_offset = 0 ){
int data_size = data.size();
preds.resize(data_size);
for(int j = 0; j < data_size; j++){
preds[j] = mparam.PredTransform(mparam.base_score +
base_model.Predict(data.data[j],buffer_index_offset + j));
}
}
private:
/*! \brief get the first order and second order gradient, given the transformed predictions and labels*/
inline void Gradient(const std::vector<float> &preds, const std::vector<float> &labels, std::vector<float> &grad,
std::vector<float> &hess){
grad.clear();
hess.clear();
for(int j = 0; j < preds.size(); j++){
grad.push_back(mparam.FirstOrderGradient(preds[j],labels[j]));
hess.push_back(mparam.SecondOrderGradient(preds[j],labels[j]));
}
}
enum LOSS_TYPE_LIST{
LINEAR_SQUARE,
LOGISTIC_NEGLOGLIKELIHOOD,
};
/*! \brief training parameter for regression */
struct ModelParam{
/* \brief global bias */
float base_score;
/* \brief type of loss function */
int loss_type;
ModelParam( void ){
base_score = 0.5f;
loss_type = 0;
@ -110,7 +187,7 @@ namespace xgboost{
*/
inline float FirstOrderGradient( float predt, float label ) const{
switch( loss_type ){
case 0: return predt - label;
case LINEAR_SQUARE: return predt - label;
case 1: return predt - label;
default: utils::Error("unknown loss_type"); return 0.0f;
}
@ -123,11 +200,53 @@ namespace xgboost{
*/
inline float SecondOrderGradient( float predt, float label ) const{
switch( loss_type ){
case 0: return 1.0f;
case 1: return predt * ( 1 - predt );
case LINEAR_SQUARE: return 1.0f;
case LOGISTIC_NEGLOGLIKELIHOOD: return predt * ( 1 - predt );
default: utils::Error("unknown loss_type"); return 0.0f;
}
}
/*!
* \brief calculating the loss, given the predictions, labels and the loss type
* \param preds the given predictions
* \param labels the given labels
* \return the specified loss
*/
inline float Loss(const std::vector<float> &preds, const std::vector<float> &labels) const{
switch( loss_type ){
case LINEAR_SQUARE: return SquareLoss(preds,labels);
case LOGISTIC_NEGLOGLIKELIHOOD: return NegLoglikelihoodLoss(preds,labels);
default: utils::Error("unknown loss_type"); return 0.0f;
}
}
/*!
* \brief calculating the square loss, given the predictions and labels
* \param preds the given predictions
* \param labels the given labels
* \return the summation of square loss
*/
inline float SquareLoss(const std::vector<float> &preds, const std::vector<float> &labels) const{
float ans = 0.0;
for(int i = 0; i < preds.size(); i++)
ans += pow(preds[i] - labels[i], 2);
return ans;
}
/*!
* \brief calculating the square loss, given the predictions and labels
* \param preds the given predictions
* \param labels the given labels
* \return the summation of square loss
*/
inline float NegLoglikelihoodLoss(const std::vector<float> &preds, const std::vector<float> &labels) const{
float ans = 0.0;
for(int i = 0; i < preds.size(); i++)
ans -= labels[i] * log(preds[i]) + ( 1 - labels[i] ) * log(1 - preds[i]);
return ans;
}
/*!
* \brief transform the linear sum to prediction
* \param x linear sum of boosting ensemble
@ -135,11 +254,13 @@ namespace xgboost{
*/
inline float PredTransform( float x ){
switch( loss_type ){
case 0: return x;
case 1: return 1.0f/(1.0f + expf(-x));
case LINEAR_SQUARE: return x;
case LOGISTIC_NEGLOGLIKELIHOOD: return 1.0f/(1.0f + expf(-x));
default: utils::Error("unknown loss_type"); return 0.0f;
}
}
};
private:
booster::GBMBaseModel base_model;
@ -147,8 +268,9 @@ namespace xgboost{
const DMatrix *train_;
std::vector<const DMatrix *> evals_;
std::vector<std::string> evname_;
bool silent;
};
};
}
};
#endif

2
regression/xgboost_reg_main.cpp
View File

@ -1,4 +1,4 @@
#include"xgbooost_reg_train.h"
#include"xgboost_reg_train.h"
#include"xgboost_reg_test.h"
using namespace xgboost::regression;

12
regression/xgboost_reg_test.h
View File

@ -12,8 +12,20 @@
using namespace xgboost::utils;
namespace xgboost{
namespace regression{
/*!
* \brief wrapping the testing process of the gradient
boosting regression model,given the configuation
* \author Kailong Chen: chenkl198812@gmail.com
*/
class RegBoostTest{
public:
/*!
* \brief to start the testing process of gradient boosting regression
* model given the configuation, and finally save the prediction
* results to the specified paths.
* \param config_path the location of the configuration
* \param silent whether to print feedback messages
*/
void test(char* config_path,bool silent = false){
reg_boost_learner = new xgboost::regression::RegBoostLearner(silent);
ConfigIterator config_itr(config_path);

47
regression/xgboost_reg_train.h
View File

@ -10,10 +10,23 @@
#include"../utils/xgboost_string.h"
using namespace xgboost::utils;
namespace xgboost{
namespace regression{
/*!
* \brief wrapping the training process of the gradient
boosting regression model,given the configuation
* \author Kailong Chen: chenkl198812@gmail.com
*/
class RegBoostTrain{
public:
/*!
* \brief to start the training process of gradient boosting regression
* model given the configuation, and finally saved the models
* to the specified model directory
* \param config_path the location of the configuration
* \param silent whether to print feedback messages
*/
void train(char* config_path,bool silent = false){
reg_boost_learner = new xgboost::regression::RegBoostLearner(silent);
ConfigIterator config_itr(config_path);
@ -39,28 +52,31 @@ namespace xgboost{
//begin training
reg_boost_learner->InitTrainer();
char model_path[256];
char suffix[256];
for(int i = 1; i <= train_param.boost_iterations; i++){
reg_boost_learner->UpdateOneIter(i);
//save the models during the iterations
if(train_param.save_period != 0 && i % train_param.save_period == 0){
sscanf(model_path,"%s/%d.model",train_param.model_dir_path,i);
FILE* file = fopen(model_path,"w");
FileStream fin(file);
reg_boost_learner->SaveModel(fin);
fin.Close();
sscanf(suffix,"%d.model",i);
SaveModel(suffix);
}
}
//save the final model
sscanf(model_path,"%s/final.model",train_param.model_dir_path);
FILE* file = fopen(model_path,"w");
FileStream fin(file);
reg_boost_learner->SaveModel(fin);
fin.Close();
//save the final round model
SaveModel("final.model");
}
private:
/*! \brief save model in the model directory with specified suffix*/
void SaveModel(const char* suffix){
char model_path[256];
//save the final round model
sscanf(model_path,"%s/%s",train_param.model_dir_path,suffix);
FILE* file = fopen(model_path,"w");
FileStream fin(file);
reg_boost_learner->SaveModel(fin);
fin.Close();
}
struct TrainParam{
/* \brief upperbound of the number of boosters */
int boost_iterations;
@ -99,7 +115,10 @@ namespace xgboost{
}
};
/*! \brief the parameters of the training process*/
TrainParam train_param;
/*! \brief the gradient boosting regression tree model*/
xgboost::regression::RegBoostLearner* reg_boost_learner;
};
}

7
regression/xgboost_regdata.h
View File

@ -30,6 +30,13 @@ namespace xgboost{
public:
/*! \brief default constructor */
DMatrix( void ){}
/*! \brief get the number of instances */
inline int size() const{
return labels.size();
}
/*!
* \brief load from text file
* \param fname name of text data

1
utils/xgboost_config.h
View File

@ -10,6 +10,7 @@
#include <cstring>
#include <string>
#include "xgboost_utils.h"
#include <vector>
namespace xgboost{
namespace utils{