vectorise_persistence#

vectorise_persistence(feature_persistence, method='statistics')#

Vectorise feature peristences.

Parameters:
feature_persistencedict

A dictionary containing persistence diagrams. The key ‘dgms’ should map to a list of arrays, where each array corresponds to a homology dimension and contains pairs of birth and death times.

methodstr, optional

The method to use for vectorising the persistence diagram. Defaults to ‘statistics’. Currently, only ‘statistics’ is implemented.

Returns:
vectorised_persistencenumpy.ndarray

A vectorised representation of the persistence diagram. The length and content of the vector depend on the chosen method.

feature_namelist

A list of the statistics corresponding to each entry of the returned vector

Raises:
ValueError

If the specified method is not implemented.

Notes

Implements vectorisation methods reviewed in https://arxiv.org/pdf/2212.09703

The ‘statistics’ method calculates the following statistics for each homology dimension:
  • Number of points in the diagram

  • Mean, std, 10th, 25th, 50th, 75th, and 90th percentiles of the birth, death, persistence and bar midpoint values

  • Entropy of the diagram

The vectorised representation is a concatenation of these statistics for all homology dimensions.