dowker_statistics

Tools for applying the Dowker complex statistically

This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) SafeDocs program under contract HR001119C0072. Any opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of DARPA.

Dependency notes

Most everything can be installed with install.packages from CRAN. The only exception is the relations.kappa function, which is from [https://github.com/kpewing/relations]. If not desired, the code chunk in question in nitf_demo.R can be skipped.

Workflows to be supported

Explanation workflow

Goal: Determine the "interesting" feature patterns

Tool: dowker_nest, pca_generator

Classification workflow

Goal: Find the probability that a given observation will be assigned a given class

Tool: dowker_prob, bernoulli_table

Unsupervised dialect clustering

Goal: Split the observations into non-inconsistent subsets

Tool: max_decreasing_decomp

Example use cases

• CSV file dialects [https://dra.american.edu/islandora/object/auislandora%3A97871]
• NITF file dialects [https://doi.org/10.1109/SPW59333.2023.00019] and [https://arxiv.org/abs/2304.09082]
• PDF files [https://doi.org/10.1109/SPW54247.2022.9833862] and [https://arxiv.org/abs/2201.08267]