#include <toolkits/supervised_learning/supervised_learning.hpp>

Public Member Functions
virtual void	train ()=0

std::vector< std::vector< flexible_type > >	get_metadata_mapping ()

virtual flexible_type	predict_single_example (const ml_data_iterator &it, const prediction_type_enum &output_type=prediction_type_enum::NA)

virtual flexible_type	predict_single_example (const DenseVector &x, const prediction_type_enum &output_type=prediction_type_enum::NA)

virtual flexible_type	predict_single_example (const SparseVector &x, const prediction_type_enum &output_type=prediction_type_enum::NA)

virtual std::map< std::string, variant_type >	evaluate (const ml_data &test_data, const std::string &evaluation_type="", bool with_prediction=false)

virtual std::map< std::string, variant_type >	evaluate (const sframe &X, const sframe &y, const std::string &evaluation_type="", bool with_prediction=false)

virtual std::shared_ptr< sarray< flexible_type > >	predict (const ml_data &test_data, const std::string &output_type="")

virtual std::shared_ptr< sarray< flexible_type > >	predict (const sframe &X, const std::string &output_type="")

virtual std::shared_ptr< sarray< flexible_type > >	extract_features (const sframe &X, ml_missing_value_action missing_value_action)

virtual sframe	predict_topk (const sframe &test_data, const std::string &output_type="", size_t topk=5)

virtual sframe	predict_topk (const ml_data &test_data, const std::string &output_type="", size_t topk=5)

virtual sframe	classify (const ml_data &test_data, const std::string &output_type="")

virtual sframe	classify (const sframe &X, const std::string &output_type="")

virtual gl_sarray	fast_predict (const std::vector< flexible_type > &rows, const std::string &missing_value_action="error", const std::string &output_type="")

virtual gl_sframe	fast_predict_topk (const std::vector< flexible_type > &rows, const std::string &missing_value_action="error", const std::string &output_type="", const size_t topk=5)

virtual gl_sframe	fast_classify (const std::vector< flexible_type > &rows, const std::string &missing_value_action="error")

virtual void	init (const sframe &X, const sframe &y, const sframe &valid_X=sframe(), const sframe &valid_y=sframe(), ml_missing_value_action mva=ml_missing_value_action::ERROR)

virtual void	set_coefs (const DenseVector &coefs)

void	set_evaluation_metric (std::vector< std::string > _metrics)

void	set_tracking_metric (std::vector< std::string > _metrics)

void	set_more_warnings (bool more_warnings)

virtual void	set_default_evaluation_metric ()

virtual void	set_default_tracking_metric ()

std::map< std::string, flexible_type >	get_train_stats () const

sframe	impute_missing_columns_using_current_metadata (const sframe &X) const

ml_data	construct_ml_data_using_current_metadata (const sframe &X, const sframe &y, ml_missing_value_action mva=ml_missing_value_action::ERROR) const

ml_data	construct_ml_data_using_current_metadata (const sframe &X, ml_missing_value_action mva=ml_missing_value_action::ERROR) const

size_t	num_features () const

size_t	num_examples () const

size_t	num_unpacked_features () const

std::vector< std::string >	get_feature_names () const

std::string	get_target_name () const

std::shared_ptr< ml_metadata >	get_ml_metadata () const

virtual bool	is_classifier () const =0

bool	is_dense ()

std::vector< std::string >	get_metrics () const

std::vector< std::string >	get_tracking_metrics () const

std::string	get_metric_display_name (const std::string &metric) const

void	display_regression_training_summary (std::string model_display_name) const

void	display_classifier_training_summary (std::string model_display_name, bool simple_mode=false) const

virtual void	model_specific_init (const ml_data &data, const ml_data &validation_data)

virtual bool	support_missing_value () const

void	api_train (gl_sframe data, const std::string &target, const variant_type &validation_data, const std::map< std::string, flexible_type > &_options)

gl_sarray	api_predict (gl_sframe data, std::string missing_value_action, std::string output_type)

gl_sframe	api_predict_topk (gl_sframe data, std::string missing_value_action, std::string output_type, size_t topk=5)

gl_sframe	api_classify (gl_sframe data, std::string missing_value_action, std::string output_type)

variant_map_type	api_evaluate (gl_sframe data, std::string missing_value_action, std::string metric, gl_sarray predictions=gl_sarray(), bool with_prediction=false)

gl_sarray	api_extract_features (gl_sframe data, std::string missing_value_action)

virtual std::shared_ptr< coreml::MLModelWrapper >	export_to_coreml ()=0

virtual void	init_options (const std::map< std::string, flexible_type > &_options)

std::vector< std::string >	list_fields ()

const variant_type &	get_value_from_state (std::string key)

const std::map< std::string, flexible_type > &	get_current_options () const

std::map< std::string, flexible_type >	get_default_options () const

const flexible_type &	get_option_value (const std::string &name) const

const std::map< std::string, variant_type > &	get_state () const

bool	is_trained () const

void	set_options (const std::map< std::string, flexible_type > &_options)

void	add_or_update_state (const std::map< std::string, variant_type > &dict)

const std::vector< option_handling::option_info > &	get_option_info () const

virtual const char *	name ()=0

virtual const std::string &	uid ()=0

virtual void	save_impl (oarchive &oarc) const

virtual void	load_version (iarchive &iarc, size_t version)

void	save_to_url (const std::string &url, const variant_map_type &side_data={})

void	save_model_to_data (std::ostream &out)

virtual size_t	get_version () const

const std::map< std::string, std::vector< std::string > > &	list_functions ()

const std::vector< std::string > &	list_get_properties ()

const std::vector< std::string > &	list_set_properties ()

variant_type	call_function (const std::string &function, variant_map_type argument)

variant_type	get_property (const std::string &property)

variant_type	set_property (const std::string &property, variant_map_type argument)

const std::string &	get_docstring (const std::string &symbol)

virtual void	perform_registration ()

Protected Member Functions
void	register_function (std::string fnname, const std::vector< std::string > &arguments, impl_fn fn)

void	register_defaults (const std::string &fnname, const variant_map_type &arguments)

void	register_setter (const std::string &propname, impl_fn setfn)

void	register_getter (const std::string &propname, impl_fn getfn)

void	register_docstring (const std::pair< std::string, std::string > &fnname_docstring)

Protected Attributes
std::map< std::string, variant_type >	state

Detailed Description

Supervised_learning model base class.

Base class for handling supervised learning class. This class is meant to be a guide to aid model writing and not a hard and fast rule of how the code must be structured.

Each supervised learning C++ toolkit contains the following:

*) model: This is the key-value map that stores the "model" attributes. The value is of type "variant_type" which is fully interfaced with python. You can add basic types, vectors, SFrames etc.

*) ml_mdata: A globally consistent object with column wise metadata. This metadata changes with time (even after training). If you want to freeze the metadata after training, you have to do so yourself.

*) train_feature_size: Feature sizes (i.e column sizes) during train time. Numerical features are of size 1, categorical features are of size (# categories), vector features are of size length, and dictionary features are of size # keys.

*) options: Option manager which keeps track of default options, current options, option ranges, type etc. This must be initialized only once in the set_options() function.

Functions that should always be implemented. Here are some notes about each of these functions that may help guide you in writing your model.

*) name: Get the name of this model. You might thinks that this is silly but the name holds the key to everything. The unity_server can construct model_base objects and they can be cast to a model of this type. The name determine how the casting happens. The init_models() function in unity_server.cpp will give you an idea of how this interface happens.

*) train: A train function for the model.

*) predict_single_example: A predict function for the model for single example. If this is implemented, batch predictions and evaluation need not be implemented.

*) predict: A predict function for the model for batch predictions. The result of this function can be an SArray of predictions. One for each value of the input SFrame.

*) evaluate: An evaluattion function for the model for evaluations. The result of this function must be an updated evaluation_stats map which can be queried with the get_evaluation_stats().

*) save: Save the model with the turicreate iarc. Turi is a server-client module. DO NOT SAVE ANYTHING in the client side. Make sure that everything is in the server side. For example: You might be tempted do keep options that the user provides into the server side but DO NOT do that because save and load will break things for you!

*) load: Load the model with the turicreate oarc.

*) init_options: Init the options

This class interfaces with the SupervisedLearning class in Python and works end to end once the following set of fuctions are implemented by the user.

Example Class

See the file supervised_learning_model.cxx for an example of how to use this class in building your supervised learning method.

Definition at line 179 of file supervised_learning.hpp.

Member Function Documentation

◆ add_or_update_state()

void turi::ml_model_base::add_or_update_state ( const std::map< std::string, variant_type > & dict )

inherited

Append the key value store of the model.

Parameters

[in] dict Options (Key-Value pairs) to set