This R package extends Jeff Scargle's Bayesian Blocks, an algorithm for optimal piecewise segmentation of non-homogeneous Poisson process data. The extensions allow for block segmentations to have non-constant intensities. Depending on the assumptions of the dataset, block segmentations can be allowed to take on power and exponential function shapes.
Special thanks to Dr. Jeff Scargle at NASA Ames and Dr. Brad Jackson at SJSU.
To play around with a demo of how this algorithm works, click the following link for an R Shiny web app. https://rshiroma.shinyapps.io/bayesianblocks2/
See pages 62-84 titled "Generalized Block Shapes" in our draft paper
https://github.com/ryanshiroma/BayesianBlocks-RPackage/raw/master/camcos2017draft.pdf