medial::process Namespace Reference

process namespace More...

Functions
void	split_feature_to_bins (const BinSettings &setting, vector< float > &feature, const vector< int > &sel_indexes, vector< float > &y)
	splits feature to bin using setting
void	normalize_feature_to_uniform (const BinSettings &setting, vector< float > &feature)
	normalize feature to be between [0-1] but also change the distribution of values to be uniform
void	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.
template<class T >
void	commit_selection (vector< T > &vec, const vector< int > &idx)
	commit selection of indexes on vector
void	filter_row_indexes (MedFeatures &dataMat, vector< int > &selected_indexes, bool op_flag=false)
	filtering MedFeatures by selected indexes rows
void	filter_row_indexes_safe (MedFeatures &dataMat, const vector< int > &selected_indexes, bool op_flag=false)
	filtering MedFeatures by selected indexes rows (thread safe for selected_indexes) no sort of selected_indexes
void	down_sample (MedFeatures &dataMat, double take_ratio, bool with_repeats=false, vector< int > *selected_indexes=NULL)
	down sampling with ratio
double	reweight_by_general (MedFeatures &data_records, const vector< string > &groups, vector< float > &weigths, bool print_verbose)
	reweighting method by given groups uniq values. return weights and min_factor
void	match_by_general (MedFeatures &data_records, const vector< string > &groups, vector< int > &filtered_row_ids, float price_ratio, int min_grp_size, bool print_verbose)
	matching by given groups uniq values. returns also the row_ids filtered
void	match_by_general (MedFeatures &data_records, const vector< string > &groups, vector< int > &filtered_row_ids, float price_ratio, float max_ratio, int min_grp_size, bool print_verbose)
	matching by given groups uniq values. returns also the row_ids filtered. max_ratio is maximal allowed ratio, inf if < 0.
void	split_matrix (const MedFeatures &matrix, vector< int > &folds, int iFold, MedFeatures &trainMatrix, MedFeatures &testMatrix, const vector< string > *selected_features=NULL)
	split matrix to train test based on iFold value. folds is fold id for each sample
void	split_matrix (const MedFeatures &matrix, unordered_map< int, int > &folds, int iFold, MedFeatures &trainMatrix, MedFeatures &testMatrix, const vector< string > *selected_features=NULL)
	split matrix to train test based on iFold value. folds is map from patient id to fold
void	convert_prctile (vector< float > &features_prctiles)
	convert feature vector to it's prctil's value in each element
void	match_to_prior (const vector< float > &outcome, const vector< float > &group_values, float target_prior, vector< int > &sel_idx)
	does matching to specific target_prior.
double	match_to_prior (MedSamples &samples, float target_prior, vector< int > &sel_idx)
	does matching to specific prior for MedSamples
double	match_to_prior (MedFeatures &features, float target_prior, vector< int > &sel_idx)
	does matching to specific prior for MedFeatures
void	match_to_prior (MedFeatures &features, const vector< string > &group_values, float target_prior, vector< int > &sel_idx, bool print_verbose=true)
	does matching to specific target_prior.
int	nSplits (vector< MedSample > &samples)
	Return number of splits, also check mismatches between idSample and internal MedSamples and set idSamples.split if missing.
float	match_multi_class (MedFeatures &data, const vector< string > &groups, vector< int > &filtered_row_ids, vector< float > &price_ratios, int nRand=10000, int verbose=false)
	multi-class matching.
float	match_multi_class (vector< MedSample > &data, const vector< string > &groups, vector< int > &filtered_row_ids, vector< float > &price_ratios, int nRand=10000, int verbose=false)
void	match_multi_class_to_dist (MedFeatures &data, const vector< string > &groups, vector< int > &filtered_row_ids, vector< float > probs)
void	match_multi_class_to_dist (vector< MedSample > &data, const vector< string > &groups, vector< int > &filtered_row_ids, vector< float > probs)
void	down_sample (MedSamples &samples, double take_ratio, bool with_repeats=false)
	down sammling
void	down_sample_by_pid (MedSamples &samples, double take_ratio, bool with_repeats=false)
	down sample by selecting from pids
void	down_sample (MedSamples &samples, int no_more_than, bool with_repeats=false)
	down sammling
void	down_sample_by_pid (MedSamples &samples, int no_more_than, bool with_repeats=false)
	down sample by selecting from pids
template<class T >
void	prctils (const vector< T > &x, const vector< double > &prc, vector< T > &res, const vector< float > *weights=NULL)
	calc prctile
template<typename T >
int	binary_search_index (const T *begin, const T *end, T val)
	binary search for index. -1 if not found
template<typename T >
int	binary_search_position (const T *begin, const T *end, T val, bool reversed=false)
	binary search for position to add new element in sorted manner (first position if equal elements found).
template<typename T >
int	binary_search_position_last (const T *begin, const T *end, T val, bool reversed=false)
	binary search for position to add new element in sorted manner (last position if equal elements found).
template<class T >
int	binary_search_position (const vector< T > &v, T search)
	binary search for position to add new element in sorted manner (first position if equal elements found).
template<class T >
int	binary_search_position (const vector< T > &v, T search, int start, int end)
	binary search for position to add new element in sorted manner (first position if equal elements found).

Detailed Description

process namespace

process

Function Documentation

compare_populations()

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  )

protected

compares two matrixes populations.

it's also try to seperate between populations using the predictor parameters if given

match_to_prior() [1/2]

void medial::process::match_to_prior	(	const vector< float > &	outcome,
		const vector< float > &	group_values,
		float	target_prior,
		vector< int > &	sel_idx
	)

does matching to specific target_prior.

Parameters

outcome	is the outcome vector for measure prior in each group
group_values	is the groups to split the matching to. it can be year signature or age or unique combination of both
target_prior	the target prior
the	return value of selected indexes to do the matching

match_to_prior() [2/2]

void medial::process::match_to_prior	(	MedFeatures &	features,
		const vector< string > &	group_values,
		float	target_prior,
		vector< int > &	sel_idx,
		bool	print_verbose = true
	)

does matching to specific target_prior.

Parameters

features	the matrix to match. will use outcome in samples
group_values	is the groups to split the matching to. it can be year signature or age or unique combination of both
target_prior	the target prior
sel_idx	the original indecies

split_feature_to_bins()

void medial::process::split_feature_to_bins	(	const BinSettings &	setting,
		vector< float > &	feature,
		const vector< int > &	sel_indexes,
		vector< float > &	y
	)

splits feature to bin using setting

Parameters

setting	the settings of split
feature	the feature vector values
sel_indexes	the indexes to take from feature. if empty will take all feature vector
y	labels if we have. some binning methods uses the labels for better split

Returns

it updates feature to the binned values