feature_filter

Module: missions.nodes.feature_selection.feature_filter

Reduce filters with the help of name filters

Inheritance diagram for pySPACE.missions.nodes.feature_selection.feature_filter:

Inheritance diagram of pySPACE.missions.nodes.feature_selection.feature_filter

FeatureNameFilterNode

class pySPACE.missions.nodes.feature_selection.feature_filter.FeatureNameFilterNode(exclude_names=[], include_names=[], exclude_patterns=[], include_patterns=[], filter_indices=[], **kwargs)[source]

Bases: pySPACE.missions.nodes.base_node.BaseNode

Filter feature vectors by name patterns or indices

Parameters

One of the following three filters should be specified.

filter_indices:

If you now the indices of features you want to delete, use this parameter. Numbering begins with zero.

(optional, default: [])

exclude_patterns:
 

List of exclude patterns

Each entry is checked, if it is included in one of the feature names. If this is the case it is deleted. In the other case it is kept.

A special functionality comes by using All instead of a list. This means to exclude everything (except the include patterns).

(optional, default: [])

exclude_names:

Feature names to be excluded

(optional, default: [])

include_patterns:
 

Skip excluded feature name fulfilling the one of these rules

(optional, default: [])

include_names:

Feature names to be included (more priority than exclusion)

(optional, default: [])

Priority Ranking

  1. include_names
  2. exclude_names
  3. include_patterns
  4. exclude_patterns
  5. filter_indices

Exemplary Call

-
    node : FeatureNameFilterNode
    parameters :
          filter_indices : [-1]
          exclude: [EMG, EOG]
          include: [EMGCLassifier, Pulse]
Input:

FeatureVector

Output:

FeatureVector

Author:

Mario Krell (mario.krell@dfki.de)

Date:

2012/08/24

POSSIBLE NODE NAMES:
 
  • FeatureNameFilter
  • FeatureNameFilterNode
POSSIBLE INPUT TYPES:
 
  • FeatureVector

Class Components Summary

_execute(data) Construct filter at first call and apply it on every vector
build_feature_selector(data) Define the retained_channel_indices for final projection
input_types
__init__(exclude_names=[], include_names=[], exclude_patterns=[], include_patterns=[], filter_indices=[], **kwargs)[source]
_execute(data)[source]

Construct filter at first call and apply it on every vector

build_feature_selector(data)[source]

Define the retained_channel_indices for final projection

input_types = ['FeatureVector']