type_conversion

Module: missions.nodes.type_manipulation.type_conversion

Convert feature to prediction vectors and TimeSeries and vice versa

Known issues
No unit tests!

Inheritance diagram for pySPACE.missions.nodes.type_manipulation.type_conversion:

Inheritance diagram of pySPACE.missions.nodes.type_manipulation.type_conversion

Class Summary

Prediction2FeaturesNode([name]) Use the prediction values as features
Features2PredictionNode(class_labels, \*\*kwargs) Use the feature vectors as prediction values
FeatureVector2TimeSeriesNode([reshape]) Convert feature vector to time series
Feature2MonoTimeSeriesNode([store, retrain, ...]) Convert feature vector to time series with only one time stamp
MonoTimeSeries2FeatureNode([store, retrain, ...]) Convert time series with only one time stamp to feature vector
CastDatatypeNode([datatype, selected_channels]) Changes the datatype of the data

Function Summary

uniquify_list(seq) Uniquify a list by preserving its original order

Classes

Prediction2FeaturesNode

class pySPACE.missions.nodes.type_manipulation.type_conversion.Prediction2FeaturesNode(name='', **kwargs)[source]

Bases: pySPACE.missions.nodes.base_node.BaseNode

Use the prediction values as features

This node converts the type PredictionVector to the type FeatureVector. This is needed, whenever one want to feed classification predictions into a node that expects feature vectors (e.g. gating functions).

Parameters

name:

String. A prefix of the new feature.

(optional, default: ‘’)

Exemplary Call

-
    node : Prediction2Features
    parameters :
        name : "SVM_"
Author:

Mario Krell (Mario.Krell@dfki.de)

Created:

2010/08/06

POSSIBLE NODE NAMES:
 
  • Prediction2Features
  • Prediction2FeaturesNode
POSSIBLE INPUT TYPES:
 
  • PredictionVector

Class Components Summary

_execute(data) Extract the prediction features from the given data
get_output_type(input_type[, as_string])
input_types
input_types = ['PredictionVector']
__init__(name='', **kwargs)[source]
_execute(data)[source]

Extract the prediction features from the given data

get_output_type(input_type, as_string=True)[source]

Features2PredictionNode

class pySPACE.missions.nodes.type_manipulation.type_conversion.Features2PredictionNode(class_labels, **kwargs)[source]

Bases: pySPACE.missions.nodes.base_node.BaseNode

Use the feature vectors as prediction values

This node converts the type PredictionVector to the type FeatureVector. The feature values are used as individual predictions and the labels are created based on the passed parameter “class_labels”.

Parameters

class_labels:

List of length two of class_labels

If a feature’s values is larger than 0, the second class label is used as the prediction vector’s label otherwise the first.

Exemplary Call

-
    node : Features2Prediction
    parameters :
        class_labels : ['Standard', 'Target']
Author:

Jan Hendrik Metzen (jhm@informatik.uni-bremen.de)

Created:

2010/089/24

POSSIBLE NODE NAMES:
 
  • Features2PredictionNode
  • Features2Prediction
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

_execute(data) Extract the prediction features from the given data
get_output_type(input_type[, as_string])
input_types
input_types = ['FeatureVector']
__init__(class_labels, **kwargs)[source]
_execute(data)[source]

Extract the prediction features from the given data

get_output_type(input_type, as_string=True)[source]

FeatureVector2TimeSeriesNode

class pySPACE.missions.nodes.type_manipulation.type_conversion.FeatureVector2TimeSeriesNode(reshape=False, **kwargs)[source]

Bases: pySPACE.missions.nodes.base_node.BaseNode

Convert feature vector to time series

This node converts the type PredictionVector to TimeSeries. The feature values are extracted and put into their respective place of sensor name and time. The sampling_frequency is also calculated.

Parameters

reshape:

Assuming, that the data is in a simple structure (the features are sorted first by sensors and second by time), a simple reshape is required and no complex iteration over all entries. This speeds up the transformation and is turned on by this parameter.

If you are unsure, just leave the parameter as it is. With the first incoming sample, the structure will be checked and if possible the parameter changed.

If the structure of your data changes, you should reset this node.

(optional, default: False)

Exemplary Call

-
    node : FeatureVector2TimeSeries
Author:

Mario Michael Krell

Created:

2011/09/23

Refactored:

2013/04/24

POSSIBLE NODE NAMES:
 
  • FeatureVector2TimeSeries
  • Feature2TimeSeries
  • LabeledFeature2TimeSeries
  • FeatureVector2TimeSeriesNode
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

_execute(data) Extract feature values from and match it to their respective sensor name and time
get_output_type(input_type[, as_string])
input_types
input_types = ['FeatureVector']
__init__(reshape=False, **kwargs)[source]
_execute(data)[source]

Extract feature values from and match it to their respective sensor name and time

get_output_type(input_type, as_string=True)[source]

Feature2MonoTimeSeriesNode

class pySPACE.missions.nodes.type_manipulation.type_conversion.Feature2MonoTimeSeriesNode(store=False, retrain=False, input_dim=None, output_dim=None, dtype=None, kwargs_warning=True, **kwargs)[source]

Bases: pySPACE.missions.nodes.base_node.BaseNode

Convert feature vector to time series with only one time stamp

This node converts the type FeatureVector to TimeSeries. No real mapping of the features to the corresponding times series place is done. Instead every feature is identified with a channel.

The purpose of this node is to enable the user to use time series nodes on feature vectors, especially on feature vectors without any time structure.

Exemplary Call

-
    node : Feature2MonoTimeSeries
Author:

Mario Krell (mario.krell@dfki.de)

Created:

2012/08/31

POSSIBLE NODE NAMES:
 
  • Feature2MonoTimeSeriesNode
  • Feature2MonoTimeSeries
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

_execute(data) Identify feature names with channel names
_invert(data) The invert function is needed for the inverse node
get_output_type(input_type[, as_string])
input_types
is_invertable() Inversion is only a mapping of names
input_types = ['FeatureVector']
_execute(data)[source]

Identify feature names with channel names

is_invertable()[source]

Inversion is only a mapping of names

_invert(data)[source]

The invert function is needed for the inverse node

get_output_type(input_type, as_string=True)[source]

MonoTimeSeries2FeatureNode

class pySPACE.missions.nodes.type_manipulation.type_conversion.MonoTimeSeries2FeatureNode(store=False, retrain=False, input_dim=None, output_dim=None, dtype=None, kwargs_warning=True, **kwargs)[source]

Bases: pySPACE.missions.nodes.type_manipulation.type_conversion.Feature2MonoTimeSeriesNode

Convert time series with only one time stamp to feature vector

This node converts the type TimeSeries to FeatureVector. Each channel is mapped to one feature.

The purpose of this node is to enable the user to use time series nodes on feature vectors. Especially on feature vectors without any time structure. Therefore this node is the back transformation from the pySPACE.missions.nodes.type_manipulation.type_conversion.Feature2MonoTimeSeriesNode

Exemplary Call

-
    node : MonoTimeSeries2Feature
Author:

Mario Krell (mario.krell@dfki.de)

Created:

2012/08/31

POSSIBLE NODE NAMES:
 
  • MonoTimeSeries2Feature
  • MonoTimeSeries2FeatureNode
POSSIBLE INPUT TYPES:
 
  • TimeSeries

Class Components Summary

_execute(data) Identify channel names with feature names
_invert(data) Irrelevant inversion introduced just for completeness
get_output_type(input_type[, as_string])
input_types
input_types = ['TimeSeries']
_execute(data)[source]

Identify channel names with feature names

_invert(data)[source]

Irrelevant inversion introduced just for completeness

get_output_type(input_type, as_string=True)[source]

CastDatatypeNode

class pySPACE.missions.nodes.type_manipulation.type_conversion.CastDatatypeNode(datatype=<type 'numpy.int16'>, selected_channels=None, **kwargs)[source]

Bases: pySPACE.missions.nodes.base_node.BaseNode

Changes the datatype of the data

Parameters
datatype:

Type to cast to.

(optional, default: “eval(numpy.float64)”)

Exemplary Call

-
    node : CastDatatype
Authors:

Hendrik Woehrle (hendrik.woehrle@dfki.de)

Created:

2012/03/29

POSSIBLE NODE NAMES:
 
  • CastDatatypeNode
  • CastDatatype
POSSIBLE INPUT TYPES:
 
  • TimeSeries

Class Components Summary

_execute(data) Apply the cast
get_output_type(input_type[, as_string])
input_types
input_types = ['TimeSeries']
__init__(datatype=<type 'numpy.int16'>, selected_channels=None, **kwargs)[source]
_execute(data)[source]

Apply the cast

get_output_type(input_type, as_string=True)[source]

Function

uniquify_list()

pySPACE.missions.nodes.type_manipulation.type_conversion.uniquify_list(seq)[source]

Uniquify a list by preserving its original order