TQRF_Params Class Reference

Inheritance diagram for TQRF_Params:

Public Member Functions
int	init (map< string, string > &map)
	extra param for use when debugging
Public Member Functions inherited from SerializableObject
virtual int	version () const
	Relevant for serializations.
virtual string	my_class_name () const
	For better handling of serializations it is highly recommended that each SerializableObject inheriting class will implement the next method.
virtual void	serialized_fields_name (vector< string > &field_names) const
	The names of the serialized fields.
virtual void *	new_polymorphic (string derived_name)
	for polymorphic classes that want to be able to serialize/deserialize a pointer * to the derived class given its type one needs to implement this function to return a new to the derived class given its type (as in my_type)
virtual void	pre_serialization ()
virtual void	post_deserialization ()
virtual size_t	get_size ()
	Gets bytes sizes for serializations.
virtual size_t	serialize (unsigned char *blob)
	Serialiazing object to blob memory. return number ob bytes wrote to memory.
virtual size_t	deserialize (unsigned char *blob)
	Deserialiazing blob to object. returns number of bytes read.
size_t	serialize_vec (vector< unsigned char > &blob)
size_t	deserialize_vec (vector< unsigned char > &blob)
virtual size_t	serialize (vector< unsigned char > &blob)
virtual size_t	deserialize (vector< unsigned char > &blob)
virtual int	read_from_file (const string &fname)
	read and deserialize model
virtual int	write_to_file (const string &fname)
	serialize model and write to file
virtual int	read_from_file_unsafe (const string &fname)
	read and deserialize model without checking version number - unsafe read
int	init_from_string (string init_string)
	Init from string.
int	init_params_from_file (string init_file)
int	init_param_from_file (string file_str, string &param)
int	update_from_string (const string &init_string)
virtual int	update (map< string, string > &map)
	Virtual to update object from parsed fields.
virtual string	object_json () const

Data Fields
string	init_string = ""
string	samples_time_unit = "Date"
	sometimes it helps to keep it for debugging
int	samples_time_unit_i
int	ncateg = 2
	calculated upon init
string	time_slice_unit = "Days"
	number of categories (1 for regression)
int	time_slice_unit_i
int	time_slice_size = -1
	calculated upon init
int	n_time_slices = 1
	the size of the basic time slice, -1: is like infinity: a single time slice like a regular QRF
vector< int >	time_slices = {}
	if time_slices vector is not given, one will be created using time_slice_size and this parameter.
vector< float >	time_slices_wgts ={}
	if not empty: defines the borders of all the time lines. Enables a very flexible time slicing strategy
int	censor_cases = 0
	default is all 1.0 , but can be assigned by the user, will be used to weight the scores from different time windows
int	max_q = 200
	when calclating the time slices distributions we have an option to NOT count the preciding 0's of non 0 cases.
string	tree_type = ""
	maximal quantization
int	tree_type_i = -1
	options: regression, entropy, logrank
int	ntrees = 50
	tree type code : calulated from tree_type the string
int	max_depth = 100
	number of trees to learn
int	min_node_last_slice = 10
	maximal depth of tree
int	min_node = 10
	stopping criteria : minimal number of samples in a node in the last time slice
float	random_split_prob = 0
	stopping criteria : minimal number of samples in a node in the first time slice
int	ntry = -1
	at this probability we will split a node in a random manner, in order to add noise to the tree.
float	ntry_prob = (float)0.1
	-1: use the ntry_prob rule, > 0 : choose this number of features.
int	nsplits = -1
	choose ntry_prob * nfeat features each time
int	max_node_test_samples = 50000
	-1: check all splits for each feature , then split the max, > 0: choose this number of split points at random and choose best
int	single_sample_per_pid = 1
	when a node is bigger than this number : choose this number of random samples to make decisions
int	bag_with_repeats = 1
	when bagging select a single sample per pid (which in itself can be repeated)
float	bag_prob = (float)0.5
	weather to bag with repeats or not
float	bag_ratio = -1
	random choice of samples for each tree prob
float	bag_feat = (float)1.0
	control ratio of #0 : #NonZero of labels, if < -1 , leave as is.
int	qpoints_per_split = 0
	proportion of random features chosen for each tree
int	nvals_for_categorial = 0
	if > 0 : will only choose this random number of points to test split points at, otherwise will test all of them
vector< string >	categorial_str
	features with number of different values below nvals_for_categ will be assumed categorial
vector< string >	categorial_tags
	all features containing one of the strings defined here in their name will be assumed categorial
float	missing_val = MED_MAT_MISSING_VALUE
	all features containing these tags will be assumed categorial
string	missing_method_str = "median"
	missing value
int	missing_method = -1
	how to handle missing values: median , left, right, mean, rand
int	test_for_inf = 1
	to be initialized from missing_method_str
int	test_for_missing = 0
	will fail on non finite values in input data
int	only_this_categ = -1
	will fail if missing value found in data
int	predict_from_slice = -1
	relavant only to categorial predictions: -1: give all categs, 0 and above: give only those categs remember that currently 0 is a special category in TQRF : the control category (nothing happens, healthy, etc...)
int	predict_to_slice = -1
	will give predictions for slices [predict_from_slice,...,predict_to_slice]. if negative: all slices.
int	predict_sum_times = 0
float	case_wgt = 1
	will sum predictions over different times
int	nrounds = 1
	the weight to use for cases with y!=0 in a weighted case
float	min_p = (float)0.01
	a single round means simply running TQRF as defined with no boosting applied
float	max_p = (float)0.99
	minimal probability to trim to when recalculating weights
float	alpha = 1
	maximal probability to trip to when recalculating weights
float	wgts_pow = 2
	shrinkage factor
float	tuning_size = 0
	power for the pow(-log(p), wgts_pow) used for adaboost weights
int	tune_max_depth = 0
	size of group to tune tree weights by.
int	tune_min_node_size = 0
	max depth of a node to get a weight for. 0 means 1 weight per tree.
float	gd_rate = (float)0.01
	min node size for a node to have a weight
int	gd_batch = 1000
	gradient descent step size
float	gd_momentum = (float)0.95
	gradient descent batch size
float	gd_lambda = 0
	gradient descent momentum
int	gd_epochs = 0
	regularization
int	verbosity = 0
	0 : stop automatically , Otherwise: do this number of epochs
int	ids_to_print = 30
	for debug prints
int	debug = 0
	control debug prints in certain places
vector< double >	log_table

Member Function Documentation

init()

int TQRF_Params::init ( map< string, string > &  map )

virtual

extra param for use when debugging

initialization from string

Reimplemented from SerializableObject.

---

The documentation for this class was generated from the following files:

• Internal/TQRF/TQRF/TQRF.h
• Internal/TQRF/TQRF/TQRF.cpp