MedAlgo.h File Reference

MedAlgo - APIs to different algorithms: Linear Models, RF, GBM, KNN, and more. More...

#include <Logger/Logger/Logger.h>
#include <MedUtils/MedUtils/MedUtils.h>
#include <MedStat/MedStat/MedStat.h>
#include <MedFeat/MedFeat/MedFeat.h>
#include <QRF/QRF/QRF.h>
#include <micNet/micNet/micNet.h>
#include <string.h>
#include <limits.h>
#include <MedProcessTools/MedProcessTools/MedProcessUtils.h>
#include <SerializableObject/SerializableObject/SerializableObject.h>
#include <TQRF/TQRF/TQRF.h>
#include "svm.h"
#include <unordered_map>
#include <random>
#include <map>
#include <string>

Go to the source code of this file.

Data Structures
class	MedPredictor
	Base Interface for predictor. More...

Namespaces
namespace	medial
	medial namespace for function
namespace	medial::models
	models namespace
namespace	medial::process
	process namespace

Macros
#define	NEW_COMPLIER   false

Enumerations
enum	MedPredictorTypes {   MODEL_LINEAR_MODEL = 0 , MODEL_QRF = 1 , MODEL_KNN = 3 , MODEL_BP = 4 ,   MODEL_MARS = 5 , MODEL_GD_LINEAR = 6 , MODEL_MULTI_CLASS = 7 , MODEL_XGB = 8 ,   MODEL_LASSO = 9 , MODEL_MIC_NET = 10 , MODEL_BOOSTER = 11 , MODEL_DEEP_BIT = 12 ,   MODEL_LIGHTGBM = 13 , MODEL_SPECIFIC_GROUPS_MODELS = 14 , MODEL_SVM = 15 , MODEL_LINEAR_SGD = 16 ,   MODEL_VW = 17 , MODEL_TQRF = 18 , MODEL_BART = 19 , MODEL_EXTERNAL_NN = 20 ,   MODEL_SIMPLE_ENSEMBLE = 21 , MODEL_BY_MISSING_VALUES_SUBSET = 22 , MODEL_LAST }

Functions
MedPredictorTypes	predictor_name_to_type (const string &model_name)
	Maping from model name in string to enum MedPredictorTypes.
int	KMeans (MedMat< float > &x, int K, MedMat< float > &centers, vector< int > &clusters, MedMat< float > &dists)
	K-Means: x is input matrix(each row is sample N*M).
int	KMeans (MedMat< float > &x, int K, int max_iter, MedMat< float > &centers, vector< int > &clusters, MedMat< float > &dists)
	K-Means: x is input matrix(each row is sample N*M).
int	KMeans (float *x, int nrows, int ncols, int K, float *centers, int *clusters, float *dists)
	K-Means: x is input matrix(each row is sample N*M).
int	KMeans (float *x, int nrows, int ncols, int K, int max_iter, float *centers, int *clusters, float *dists, bool verbose_print=true)
	K-Means: x is input matrix(each row is sample N*M).
int	MedPCA (MedMat< float > &x, MedMat< float > &pca_base, vector< float > &varsum)
	given a matrix, returns the base PCA matrix and the cummulative relative variance explained by them.
int	MedPCA_project (MedMat< float > &x, MedMat< float > &pca_base, int dim, MedMat< float > &projected)
	returns the projection of the pca base on the first dim dimensions.
string	medial::models::getParamsInfraModel (void *model)
	returns string to create model with init_string. void * is MedPredictor
void *	medial::models::copyInfraModel (void *model, bool delete_old=true)
	returns MedPredictor *, a clone copy of given model (params without learned data). if delete_old is true will free old given model
void	medial::models::initInfraModel (void *&model)
	initialize model which is MedPredictor by copying it's parameters to new address and freeing old one
void	medial::models::learnInfraModel (void *model, const vector< vector< float > > &xTrain, vector< float > &y, vector< float > &weights)
	run Learn on the MedPredictor - wrapper api
vector< float >	medial::models::predictInfraModel (void *model, const vector< vector< float > > &xTest)
	run predict on the MedPredictor - wrapper api
void	medial::models::get_pids_cv (MedPredictor *pred, MedFeatures &matrix, int nFolds, mt19937 &generator, vector< float > &preds)
	run cross validation where each pid is in diffrent fold and saves the preds.
void	medial::models::get_cv (MedPredictor *pred, MedFeatures &matrix, int nFolds, mt19937 &generator, vector< float > &preds)
	run cross validation where each samples can be in diffrent fold and saves the preds.
void	medial::process::compare_populations (const MedFeatures &population1, const MedFeatures &population2, const string &name1, const string &name2, const string &output_file, const string &predictor_type="", const string &predictor_init="", int nfolds=5, int max_learn=0)
	compares two matrixes populations.

Variables
unordered_map< int, string >	predictor_type_to_name
	Maping from predictor enum type MedPredictorTypes to model name in string.

Detailed Description

MedAlgo - APIs to different algorithms: Linear Models, RF, GBM, KNN, and more.

Enumeration Type Documentation

MedPredictorTypes

enum MedPredictorTypes

Enumerator
MODEL_LINEAR_MODEL	to_use:"linear_model" Linear Model - creates MedLM
MODEL_QRF	to_use:"qrf" Q-Random-Forest - creates MedQRF
MODEL_KNN	to_use:"knn" K Nearest Neighbour - creates MedKNN
MODEL_BP	to_use:"BP" Neural Network Back Propagation - creates MedBP
MODEL_MARS	to_use:"mars" Multivariate Adaptive Regression Splines - creates MedMars
MODEL_GD_LINEAR	to_use:"gdlm" Gradient Descent/Full solution ridge - creates MedGDLM
MODEL_MULTI_CLASS	to_use:"multi_class" general one vs. all multi class extention - creates MedMultiClass
MODEL_XGB	to_use:"xgb" XGBoost - creates MedXGB
MODEL_LASSO	to_use:"lasso" Lasso model - creates MedLasso
MODEL_MIC_NET	to_use:"micNet" Home brew Neural Net implementation (Allows deep learning) - creates MedMicNet
MODEL_BOOSTER	to_use:"booster" general booster (meta algorithm) - creates MedBooster
MODEL_DEEP_BIT	to_use:"deep_bit" Nir\'s DeepBit method - creates MedDeepBit
MODEL_LIGHTGBM	to_use:"lightgbm" the celebrated LightGBM algorithm - creates MedLightGBM
MODEL_SPECIFIC_GROUPS_MODELS	to_use:"multi_models" spliting model by specific value (for example age-range) and train diffretn model for each bin - creates MedSpecificGroupModels
MODEL_SVM	to_use:"svm" Svm model - creates MedSvm
MODEL_LINEAR_SGD	to_use:"linear_sgd" linear model using our customized SGD - creates MedLinearModel
MODEL_VW	to_use:"vw" VowpalWabbit yahoo reasearch library - creates MedVW
MODEL_TQRF	to_use:"tqrf" TQRF model - creates MedTQRF
MODEL_BART	to_use:"bart" MedBART model using BART
MODEL_EXTERNAL_NN	to_use: "external_nn" , initialize a neural net using a layers file. creates MedExternalNN
MODEL_SIMPLE_ENSEMBLE	to_use: "simple_ensemble" , give 1 or more models to train, and ensemble them with given weights from the user. creates MedSimpleEnsemble
MODEL_BY_MISSING_VALUES_SUBSET	to_use: "by_missing_value_subset", choosed MedPredictor on subset of the features based on missing values. choose best fit - creates MedPredictorsByMissingValues.

Function Documentation

KMeans() [1/4]

int KMeans	(	float *	x,
		int	nrows,
		int	ncols,
		int	K,
		float *	centers,
		int *	clusters,
		float *	dists
	)

K-Means: x is input matrix(each row is sample N*M).

K- number of clusters, centers - output centroids of clusters(K*M) clusters - output for each sample the cluster number from 0 to K-1(N*1). dists - output of distance for each sample form each cluster(N*K)

KMeans() [2/4]

int KMeans	(	float *	x,
		int	nrows,
		int	ncols,
		int	K,
		int	max_iter,
		float *	centers,
		int *	clusters,
		float *	dists,
		bool	verbose_print = true
	)

K-Means: x is input matrix(each row is sample N*M).

K- number of clusters, centers - output centroids of clusters(K*M) clusters - output for each sample the cluster number from 0 to K-1(N*1). dists - output of distance for each sample form each cluster(N*K)

KMeans() [3/4]

int KMeans	(	MedMat< float > &	x,
		int	K,
		int	max_iter,
		MedMat< float > &	centers,
		vector< int > &	clusters,
		MedMat< float > &	dists
	)

K-Means: x is input matrix(each row is sample N*M).

K- number of clusters, centers - output centroids of clusters(K*M) clusters - output for each sample the cluster number from 0 to K-1(N*1). dists - output of distance for each sample form each cluster(N*K)

KMeans() [4/4]

int KMeans	(	MedMat< float > &	x,
		int	K,
		MedMat< float > &	centers,
		vector< int > &	clusters,
		MedMat< float > &	dists
	)

K-Means: x is input matrix(each row is sample N*M).

K- number of clusters, centers - output centroids of clusters(K*M) clusters - output for each sample the cluster number from 0 to K-1(N*1). dists - output of distance for each sample form each cluster(N*K)

MedPCA()

int MedPCA	(	MedMat< float > &	x,
		MedMat< float > &	pca_base,
		vector< float > &	varsum
	)

given a matrix, returns the base PCA matrix and the cummulative relative variance explained by them.

it is highly recommended to normalize the input matrix x before calling.