fab.hme.learner.init package¶
Submodules¶
fab.hme.learner.init.init_hme_bern_gate_supervised_model module¶
- class fab.hme.learner.init.init_hme_bern_gate_supervised_model.HMEBernGateModelRandomInitializer(comp_initializer, tree_depth)¶
Bases:
HMESupervisedModelInitializer
Initializer for HMESupervisedModel with random initialization.
This class create an model in which any parameters are randomly initialized using specified scale parameter, input data and the depth of gating-tree.
- Parameters:
- comp_initializerSupervisedComponentInitializer
Component-initializer instance.
- tree_depthint
Depth of gating-tree. The complete binary tree is considered.
Methods
init_model
(data)Creates an initialized model instance.
- init_model(data)¶
Creates an initialized model instance.
- Parameters:
- dataHMESupervisedData
Input data to be used for calculating initial parameter values.
- Returns:
- modelHMESupervisedModel
Initialized model instance.
- vposteriorHMEBinaryTreeVPosterior
Variational posterior for the model.
- class fab.hme.learner.init.init_hme_bern_gate_supervised_model.HMEBernGateModelWithPosteriorInitializer(comp_initializer, posterior_prob, comp_ids)¶
Bases:
HMESupervisedModelInitializer
Initializer for HMESupervisedModel with given initial posterior.
This class create an model in which any parameters are randomly initialized using specified scale parameter, input data. The structure of gating-tree is decided by a given list of component IDs (comp_ids).
- Parameters:
- comp_initializerSupervisedComponentInitializer
Component-initializer instance.
- posterior_probnp.ndarray, shape = (num_samples, num_comps)
Initial posterior distribution. The number of columns must the same as length of comp_ids.
- comp_idslist[int], size = (num_comps)
Component ID numbers. The IDs must be ones assigned by left-to-right order in a complete binary tree.
Methods
init_model
(data)Creates an initialized model instance.
- init_model(data)¶
Creates an initialized model instance.
- Parameters:
- dataHMESupervisedData
Input data to be used for calculating initial parameter values.
- Returns:
- modelHMESupervisedModel
Initialized model instance.
- vposteriorHMEBinaryTreeVPosterior
Variational posterior for the model.
fab.hme.learner.init.init_hme_logit_gate_supervised_model module¶
- class fab.hme.learner.init.init_hme_logit_gate_supervised_model.HMELogitGateModelRandomInitializer(comp_initializer, tree_depth)¶
Bases:
HMESupervisedModelInitializer
Initializer for HMESupervisedModel with random initialization.
This class create an model in which any parameters are randomly initialized using specified scale parameter, input data and the depth of gating-tree.
- Parameters:
- comp_initializerSupervisedComponentInitializer
Component-initializer instance.
- tree_depthint
Depth of gating-tree. The complete binary tree is considered.
Methods
init_model
(data)Creates an initialized model instance.
- init_model(data)¶
Creates an initialized model instance.
- Parameters:
- dataHMESupervisedData
Input data to be used for calculating initial parameter values.
- Returns:
- modelHMESupervisedModel
Initialized model instance.
- vposteriorHMELogitGateVPosterior
Variational posterior for the model.
- class fab.hme.learner.init.init_hme_logit_gate_supervised_model.HMELogitGateModelWithPosteriorInitializer(comp_initializer, posterior_prob, comp_ids)¶
Bases:
HMESupervisedModelInitializer
Initializer for HMESupervisedModel with given initial posterior.
This class create an model in which any parameters are randomly initialized using specified scale parameter, input data. The structure of gating-tree is decided by a given list of component IDs (comp_ids).
- Parameters:
- comp_initializerSupervisedComponentInitializer
Component-initializer instance.
- posterior_probnp.ndarray, shape = (num_samples, num_comps)
Initial posterior distribution. The number of columns must the same as length of comp_ids.
- comp_idslist[int], size = (num_comps)
Component ID numbers. The IDs must be ones assigned by left-to-right order in a complete binary tree.
Methods
init_model
(data)Creates an initialized model instance.
- init_model(data)¶
Creates an initialized model instance.
- Parameters:
- dataHMESupervisedData
Input data to be used for calculating initial parameter values.
- Returns:
- modelHMESupervisedModel
Initialized model instance.
- vposteriorHMEBinaryTreeVPosterior
Variational posterior for the model.
fab.hme.learner.init.init_hme_model module¶
- class fab.hme.learner.init.init_hme_model.HMESupervisedModelInitializer¶
Bases:
object
An abstract class of initializer for HMESupervisedModel’s.
Concrete classes must implement the init_model() method that creates model instance using given data and parameters specified at __init__().
Methods
init_model
(data)Creates an initialized model instance.
- abstract init_model(data)¶
Creates an initialized model instance.
- Parameters:
- dataHMESupervisedData
Input data to be used for calculating initial parameter values.
- Returns:
- modelHMESupervisedModel
Initialized model instance.
- vposteriorHMEVPosterior
Variational posterior for the model.
- class fab.hme.learner.init.init_hme_model.HMESupervisedModelWithModelInitializer(model_dict)¶
Bases:
HMESupervisedModelInitializer
Initializer for HMESupervisedModel with given information on FAB/HME supervised model.
This class create model which is restored from model_dict.
- Parameters:
- model_dictdict
Information on FAB/HME supervised model.
Methods
init_model
(data)Creates an initialized model instance.
- init_model(data)¶
Creates an initialized model instance.
- Parameters:
- dataHMESupervisedData
Input data to be used for calculating initial parameter values.
- Returns:
- modelHMESupervisedModel
Initialized model instance.
- vposteriorHMEBinaryTreeVPosterior
Variational posterior for the model.