Population Averaging
Given a representative population sample, one could simply average these graphs and use the average adjacency with the individual connectivity as features.
This transformer is intended for this case.
Population Averaging Transform
Transformation for averaging over the population The connectivity of an individual is discarded by this transformation. Instead, the precalculated mean is used and the original connectivity will be passed as features.
Warning
The original adjacency matrix of unseen data is discarded by application of this transformer!
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
adjacency_axis |
int
|
Index of the adjacency part of the input arrays. If a GraphConstructor is used with default parameters the adjacency is stored at the first dimension. |
0
|
connectivity_axis |
int
|
Index of the connectivity part of the input arrays for unseen data. If a GraphConstructor is used with default parameters the connectivity is at the second dimension. |
1
|
feature_axis |
Union[int, Tuple[int, ...]]
|
Index (or indices) of the feature part of the unseen data. |
2
|
Example
my_pipe.add(PipelineElement('PopulationAveragingTransform'))
Notes
Using the feature axis as additional feature is not implemented yet.