spatial_filtering

Module: missions.nodes.spatial_filtering.spatial_filtering

Basic methods for spatial filtering

Spatial filtering means to combine information from mostly similar sensors, distributed in space. The main principle is to define a linear filter in a training phase, which tries to combine the sensor information to erase noise and compress the relevant information.

Inheritance diagram for pySPACE.missions.nodes.spatial_filtering.spatial_filtering:

Inheritance diagram of pySPACE.missions.nodes.spatial_filtering.spatial_filtering

SpatialFilteringNode

class pySPACE.missions.nodes.spatial_filtering.spatial_filtering.SpatialFilteringNode(retained_channels=None, **kwargs)[source]

Bases: pySPACE.missions.nodes.base_node.BaseNode

Base class for spatial filters and simple channel reduction

This class is superclass for nodes that implement spatial filtering. It contains functionality that is common to all spatial filters.

It can also be used directly as node which retains just the first N channels. This is typically only reasonable if the channel ordering is somehow meaningful.

This class shall unify processing steps like:

  • execution of the linear transformation on the data
  • ranking of sensor channels dependent on the weights in the filter (done)
  • initialization of the nodes
  • and visualization, if the sensors are EEG electrodes.
Parameters
retained_channels:
 

The number of channels that are retained. If this quantity is not defined, all channels are retained.

(optional, default: None)

Exemplary Call

-  
    node : SpatialFiltering
    parameters :
          retained_channels : 8
Author:

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

Created:

2011/11/22

POSSIBLE NODE NAMES:
 
  • SpatialFiltering
  • SpatialFilteringNode
POSSIBLE INPUT TYPES:
 
  • TimeSeries

Class Components Summary

__hyperparameters
_execute(data)
get_filters()
get_own_transformation([sample])
get_sensor_ranking() Special Code for the spatial filter
input_types
__init__(retained_channels=None, **kwargs)[source]
_execute(data)[source]
get_sensor_ranking()[source]

Special Code for the spatial filter

Take a maximum number of ranking channels and add the channel weights

Channels with the highest weight are the most important.

Should work for xDAWN, CSP, FDA, PCA, ICA

get_filters()[source]
get_own_transformation(sample=None)[source]
__hyperparameters = set([NoOptimizationParameter<kwargs_warning>, NoOptimizationParameter<dtype>, NoOptimizationParameter<output_dim>, NoOptimizationParameter<retrain>, NoOptimizationParameter<input_dim>, QLogUniformParameter<retained_channels>, NoOptimizationParameter<store>])
input_types = ['TimeSeries']