ORMM

Module: missions.nodes.classification.svm_variants.ORMM

One class variants of BRMM based on separation from zero

Inheritance diagram for pySPACE.missions.nodes.classification.svm_variants.ORMM:

Inheritance diagram of pySPACE.missions.nodes.classification.svm_variants.ORMM

Class Summary

OcRmmNode([outer_boundary]) Take zero as opposite class
OcRmmPerceptronNode([outer_boundary]) Take zero as opposite class for online learning update formula
L2OcRmmPerceptronNode(\*\*kwargs) Squared loss variant of the one-class RMM Perceptron
SvddPassiveAggressiveNode([radius, version, ...]) Support Vector Data Description like Perceptron’s suggested by Crammer
UnaryPA0Node(\*\*kwargs) PA0 Node for unary classification
UnaryPA1Node(\*\*kwargs) PA1 Node for unary classification
UnaryPA2Node(\*\*kwargs) PA2 Node for unary classification

Classes

OcRmmNode

class pySPACE.missions.nodes.classification.svm_variants.ORMM.OcRmmNode(outer_boundary=False, **kwargs)[source]

Bases: pySPACE.missions.nodes.classification.svm_variants.brmm.RMM2Node, pySPACE.missions.nodes.classification.one_class.OneClassClassifierBase

Take zero as opposite class

References

main  
author Krell, M. M. and Woehrle, H.
title New one-class classifiers based on the origin separation approach
journal Pattern Recognition Letters
publisher Elsevier
doi 10.1016/j.patrec.2014.11.008
year 2015

Exemplary Call

-   node : OcRmmNode
    parameters :
        complexity : 1.0
        class_labels : ['Target']
        range : 4
Input:

FeatureVector
Output:

PredictionVector
Author:

Mario Michael Krell
Created:

2013/08/16
POSSIBLE NODE NAMES:
 
  • OcRmmNode
  • OcRmm
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

__hyperparameters
_execute(data)
_inc_train(data, label) Special wrapper needed to avoid wrong or unknown label
append_weights_and_class_factors(label) Only label zero is expected and label factor one is used
train(data, label) Forward to one class method
__init__(outer_boundary=False, **kwargs)[source]
append_weights_and_class_factors(label)[source]

Only label zero is expected and label factor one is used

train(data, label)[source]

Forward to one class method

_execute(data)[source]
_inc_train(data, label)[source]

Special wrapper needed to avoid wrong or unknown label

Mostly code copy from train method.

__hyperparameters = set([ChoiceParameter<kernel_type>, NoOptimizationParameter<linear_weighting>, BooleanParameter<squared_loss>, NoOptimizationParameter<dtype>, NoOptimizationParameter<use_list>, NoOptimizationParameter<kwargs_warning>, BooleanParameter<regression>, QLogUniformParameter<max_iterations>, NormalParameter<ratio>, NoOptimizationParameter<output_dim>, LogNormalParameter<tolerance>, UniformParameter<nu>, NoOptimizationParameter<store>, BooleanParameter<outer_boundary>, NoOptimizationParameter<input_dim>, LogNormalParameter<epsilon>, NoOptimizationParameter<retrain>, NoOptimizationParameter<debug>, QNormalParameter<offset>, NoOptimizationParameter<omega>, LogUniformParameter<complexity>, QUniformParameter<max_time>, NoOptimizationParameter<offset_factor>, LogUniformParameter<range_>, NoOptimizationParameter<keep_vectors>])

OcRmmPerceptronNode

class pySPACE.missions.nodes.classification.svm_variants.ORMM.OcRmmPerceptronNode(outer_boundary=False, **kwargs)[source]

Bases: pySPACE.missions.nodes.classification.svm_variants.brmm.RmmPerceptronNode

Take zero as opposite class for online learning update formula

References

main  
author Krell, M. M. and Woehrle, H.
title New one-class classifiers based on the origin separation approach
journal Pattern Recognition Letters
publisher Elsevier
doi 10.1016/j.patrec.2014.11.008
year 2015

Exemplary Call

-   node : OcRmmPerceptronNode
    parameters :
        complexity : 1.0
        class_labels : ['Target']
        range : 4
Input:

FeatureVector
Output:

PredictionVector
Author:

Mario Michael Krell
Created:

2014/01/02
POSSIBLE NODE NAMES:
 
  • OcRmmPerceptron
  • OcRmmPerceptronNode
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

__hyperparameters
_execute(data)
_inc_train(data, label) Special wrapper needed to avoid wrong or unknown label
train(data, label) Code copy from OneClassClassifierBase
__init__(outer_boundary=False, **kwargs)[source]
train(data, label)[source]

Code copy from OneClassClassifierBase

_inc_train(data, label)[source]

Special wrapper needed to avoid wrong or unknown label

Mostly code copy from train method.

_execute(data)[source]
__hyperparameters = set([NoOptimizationParameter<kernel_type>, NoOptimizationParameter<linear_weighting>, BooleanParameter<squared_loss>, NoOptimizationParameter<dtype>, NoOptimizationParameter<use_list>, NoOptimizationParameter<kwargs_warning>, BooleanParameter<regression>, QLogUniformParameter<max_iterations>, NormalParameter<ratio>, NoOptimizationParameter<output_dim>, NoOptimizationParameter<version>, LogNormalParameter<tolerance>, UniformParameter<nu>, NoOptimizationParameter<store>, BooleanParameter<outer_boundary>, NoOptimizationParameter<input_dim>, LogNormalParameter<epsilon>, NoOptimizationParameter<retrain>, NoOptimizationParameter<offset_factor>, QNormalParameter<offset>, NoOptimizationParameter<omega>, LogUniformParameter<complexity>, QUniformParameter<max_time>, NoOptimizationParameter<debug>, LogUniformParameter<range_>, NoOptimizationParameter<keep_vectors>])

L2OcRmmPerceptronNode

class pySPACE.missions.nodes.classification.svm_variants.ORMM.L2OcRmmPerceptronNode(**kwargs)[source]

Bases: pySPACE.missions.nodes.classification.svm_variants.ORMM.OcRmmPerceptronNode

Squared loss variant of the one-class RMM Perceptron

References

main  
author Krell, M. M. and Woehrle, H.
title New one-class classifiers based on the origin separation approach
journal Pattern Recognition Letters
publisher Elsevier
doi 10.1016/j.patrec.2014.11.008
year 2015

See also

OcRmmPerceptronNode

Exemplary Call

-   node : L2OcRmmPerceptronNode
    parameters :
        complexity : 1.0
        class_labels : ['Target']
        range : 4
Input:

FeatureVector
Output:

PredictionVector
Author:

Mario Michael Krell
Created:

2014/04/28
POSSIBLE NODE NAMES:
 
  • L2OcRmmPerceptron
  • L2OcRmmPerceptronNode
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

__hyperparameters
__init__(**kwargs)[source]
__hyperparameters = set([NoOptimizationParameter<kernel_type>, NoOptimizationParameter<linear_weighting>, NoOptimizationParameter<squared_loss>, NoOptimizationParameter<dtype>, NoOptimizationParameter<use_list>, NoOptimizationParameter<kwargs_warning>, BooleanParameter<regression>, QLogUniformParameter<max_iterations>, NormalParameter<ratio>, NoOptimizationParameter<output_dim>, NoOptimizationParameter<version>, LogNormalParameter<tolerance>, UniformParameter<nu>, NoOptimizationParameter<store>, BooleanParameter<outer_boundary>, QUniformParameter<max_time>, LogNormalParameter<epsilon>, NoOptimizationParameter<retrain>, NoOptimizationParameter<debug>, QNormalParameter<offset>, NoOptimizationParameter<omega>, LogUniformParameter<complexity>, NoOptimizationParameter<input_dim>, NoOptimizationParameter<offset_factor>, LogUniformParameter<range_>, NoOptimizationParameter<keep_vectors>])

SvddPassiveAggressiveNode

class pySPACE.missions.nodes.classification.svm_variants.ORMM.SvddPassiveAggressiveNode(radius=0, version=1, radius_opt=False, **kwargs)[source]

Bases: pySPACE.missions.nodes.classification.svm_variants.ORMM.OcRmmPerceptronNode

Support Vector Data Description like Perceptron’s suggested by Crammer

References

main  
author Crammer, K. and Dekel, O. and Keshet, J. and Shalev-Shwartz, S. and Singer, Y.
title Online Passive-Aggressive Algorithms
journal Journal of Machine Learning Research
doi 10.1016/j.patrec.2013.09.018
volume 7
pages 551 - 585
year 2006

Parameters

radius:

Maximum range parameter allowed for sphere

(optional, default: 0)
radius_opt:

Optimize the range parameter as described in reference. If no optimization is used, the radius parameter defines the used range.

(optional, default: False)
version:

Defines the handling of loss:

  • 0: hard margin with zero loss on new sample (PA0),
  • 1: soft margin with linear loss punishment (PA1),
  • 2: soft margin with squared loss punishment (PA2).

For more details refer to the given reference.

(optional, default: 1)

Exemplary Call

-   node : SvddPassiveAggressive
    parameters :
        complexity : 1.0
        class_labels : ['Target', 'REST']
        radius : 2
        radius_opt : True
        version : 1
Input:

FeatureVector
Output:

PredictionVector
Author:

Mario Michael Krell
Created:

2014/04/17
POSSIBLE NODE NAMES:
 
  • SvddPassiveAggressive
  • SvddPassiveAggressiveNode
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

__hyperparameters
_execute(data)
_train(data, class_label)
__init__(radius=0, version=1, radius_opt=False, **kwargs)[source]
_train(data, class_label)[source]
_execute(data)[source]
__hyperparameters = set([NoOptimizationParameter<kernel_type>, NoOptimizationParameter<linear_weighting>, BooleanParameter<squared_loss>, NoOptimizationParameter<dtype>, NoOptimizationParameter<use_list>, LogUniformParameter<radius>, NoOptimizationParameter<kwargs_warning>, BooleanParameter<radius_opt>, BooleanParameter<regression>, QLogUniformParameter<max_iterations>, NormalParameter<ratio>, NoOptimizationParameter<output_dim>, ChoiceParameter<version>, LogNormalParameter<tolerance>, UniformParameter<nu>, NoOptimizationParameter<store>, BooleanParameter<outer_boundary>, QUniformParameter<max_time>, LogNormalParameter<epsilon>, NoOptimizationParameter<retrain>, NoOptimizationParameter<debug>, QNormalParameter<offset>, NoOptimizationParameter<omega>, LogUniformParameter<complexity>, NoOptimizationParameter<input_dim>, NoOptimizationParameter<offset_factor>, LogUniformParameter<range_>, NoOptimizationParameter<keep_vectors>])

UnaryPA0Node

class pySPACE.missions.nodes.classification.svm_variants.ORMM.UnaryPA0Node(**kwargs)[source]

Bases: pySPACE.missions.nodes.classification.svm_variants.ORMM.SvddPassiveAggressiveNode

PA0 Node for unary classification

See also

SvddPassiveAggressiveNode

Exemplary Call

-   node : UnaryPA0
    parameters :
        complexity : 1.0
        class_labels : ['Target', 'REST']
        radius : 2
        radius_opt : True
Input:

FeatureVector
Output:

PredictionVector
Author:

Mario Michael Krell
Created:

2014/04/17
POSSIBLE NODE NAMES:
 
  • UnaryPA0
  • UnaryPA0Node
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

__hyperparameters
__init__(**kwargs)[source]
__hyperparameters = set([NoOptimizationParameter<kernel_type>, NoOptimizationParameter<linear_weighting>, BooleanParameter<squared_loss>, NoOptimizationParameter<dtype>, NoOptimizationParameter<use_list>, LogUniformParameter<radius>, NoOptimizationParameter<kwargs_warning>, BooleanParameter<radius_opt>, BooleanParameter<regression>, QLogUniformParameter<max_iterations>, NormalParameter<ratio>, NoOptimizationParameter<output_dim>, NoOptimizationParameter<version>, LogNormalParameter<tolerance>, UniformParameter<nu>, NoOptimizationParameter<store>, BooleanParameter<outer_boundary>, NoOptimizationParameter<input_dim>, LogNormalParameter<epsilon>, NoOptimizationParameter<retrain>, NoOptimizationParameter<offset_factor>, QNormalParameter<offset>, NoOptimizationParameter<omega>, LogUniformParameter<complexity>, QUniformParameter<max_time>, NoOptimizationParameter<debug>, LogUniformParameter<range_>, NoOptimizationParameter<keep_vectors>])

UnaryPA1Node

class pySPACE.missions.nodes.classification.svm_variants.ORMM.UnaryPA1Node(**kwargs)[source]

Bases: pySPACE.missions.nodes.classification.svm_variants.ORMM.SvddPassiveAggressiveNode

PA1 Node for unary classification

See also

SvddPassiveAggressiveNode

Exemplary Call

-   node : UnaryPA1
    parameters :
        complexity : 1.0
        class_labels : ['Target', 'REST']
        radius : 2
        radius_opt : True
Input:

FeatureVector
Output:

PredictionVector
Author:

Mario Michael Krell
Created:

2014/04/17
POSSIBLE NODE NAMES:
 
  • UnaryPA1
  • UnaryPA1Node
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

__hyperparameters
__init__(**kwargs)[source]
__hyperparameters = set([NoOptimizationParameter<kernel_type>, NoOptimizationParameter<linear_weighting>, BooleanParameter<squared_loss>, NoOptimizationParameter<dtype>, NoOptimizationParameter<use_list>, LogUniformParameter<radius>, NoOptimizationParameter<kwargs_warning>, BooleanParameter<radius_opt>, BooleanParameter<regression>, QLogUniformParameter<max_iterations>, NormalParameter<ratio>, NoOptimizationParameter<output_dim>, NoOptimizationParameter<version>, LogNormalParameter<tolerance>, UniformParameter<nu>, NoOptimizationParameter<store>, BooleanParameter<outer_boundary>, NoOptimizationParameter<input_dim>, LogNormalParameter<epsilon>, NoOptimizationParameter<retrain>, NoOptimizationParameter<offset_factor>, QNormalParameter<offset>, NoOptimizationParameter<omega>, LogUniformParameter<complexity>, QUniformParameter<max_time>, NoOptimizationParameter<debug>, LogUniformParameter<range_>, NoOptimizationParameter<keep_vectors>])

UnaryPA2Node

class pySPACE.missions.nodes.classification.svm_variants.ORMM.UnaryPA2Node(**kwargs)[source]

Bases: pySPACE.missions.nodes.classification.svm_variants.ORMM.SvddPassiveAggressiveNode

PA2 Node for unary classification

See also

SvddPassiveAggressiveNode

Exemplary Call

-   node : UnaryPA2
    parameters :
        complexity : 1.0
        class_labels : ['Target', 'REST']
        radius : 2
        radius_opt : True
Input:

FeatureVector
Output:

PredictionVector
Author:

Mario Michael Krell
Created:

2014/04/17
POSSIBLE NODE NAMES:
 
  • UnaryPA2
  • UnaryPA2Node
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

__hyperparameters
__init__(**kwargs)[source]
__hyperparameters = set([NoOptimizationParameter<kernel_type>, NoOptimizationParameter<linear_weighting>, BooleanParameter<squared_loss>, NoOptimizationParameter<dtype>, NoOptimizationParameter<use_list>, LogUniformParameter<radius>, NoOptimizationParameter<kwargs_warning>, BooleanParameter<radius_opt>, BooleanParameter<regression>, QLogUniformParameter<max_iterations>, NormalParameter<ratio>, NoOptimizationParameter<output_dim>, NoOptimizationParameter<version>, LogNormalParameter<tolerance>, UniformParameter<nu>, NoOptimizationParameter<store>, BooleanParameter<outer_boundary>, NoOptimizationParameter<input_dim>, LogNormalParameter<epsilon>, NoOptimizationParameter<retrain>, NoOptimizationParameter<offset_factor>, QNormalParameter<offset>, NoOptimizationParameter<omega>, LogUniformParameter<complexity>, QUniformParameter<max_time>, NoOptimizationParameter<debug>, LogUniformParameter<range_>, NoOptimizationParameter<keep_vectors>])