FLOPART

Functional Labeled Optimal Partitioning - A peak detection algorithm that allows for three types of labels.

Installation

install.packages("FLOPART")
##OR
if(require("remotes"))install.packages("remotes")
remotes::install_github("alyssajs/FLOPART)

Usage

The main driver function is FLOPART, which takes

• data_vec An integer vector of data
• weight_vec A double vector of weights for each data point
• data_count The number of data points (an integer)
• penalty The penalty value (a double)
• cost_mat A numeric vector used for output - after execution, the first half of this vector contains the cost of being in the peak state for each data point, and the second half of this vector contains the cost of being in the background state for each data point
• end_vec An integer vector of segment ends - finite positive values represent segment endspoints, infinite or negative values are used as placeholders
• mean_vec A double vector of segment means in reverse order
• intervals_mat An integer vector that stores the number of cost function pieces - that is, the number of potential changepoints. Used to analyze time complexity.
• label_starts An integer vector of indices where labels start (0-based)
• label_ends An integer vector of indices where labels end (0-based)
• label_types An integer vector of label types where -1 corresponds to a peakEnd label, 0 corresponds to a noPeaks label, and 1 corresponds to a peakStart label
• label_count The number of labels (should match the length of the label_starts, label_ends, and label_types vector) - an integer

FLOPART has an interface function to use FLOPART in R by calling the underlying C++ code.

library(FLOPART)
data_count = 6L
data_vec <- as.integer(c(1,10,10,10,10,1))
weight_vec <- as.double(rep(1, data_count))
penalty <- as.double(0)
cost_mat <- numeric(data_count * 2)
end_vec <- integer(data_count)
mean_vec <- numeric(data_count) 
intervals_mat <- integer(data_count*2)
label_starts <- 1L
label_ends <- 3L
label_types <- 0L
label_count <- 1L

result <- 
  .C("FLOPART_interface", data_vec = data_vec, weight_vec = weight_vec,
     data_count = data_count, penalty = penalty, cost_mat = cost_mat,
   end_vec = end_vec, mean_vec = mean_vec, intervals_mat = intervals_mat,
   label_starts = label_starts, label_ends = label_ends,
   label_types = label_types, label_count = label_count,
   PACKAGE="FLOPART")

costMatrix <- matrix(result[["cost_mat"]], nrow=2, ncol=data_count, byrow=TRUE)
> costMatrix
     [,1]      [,2]      [,3]     [,4]      [,5]      [,6]
[1,]    1       Inf       Inf      Inf -9.100866 -6.621371
[2,]    1 -3.876115 -6.621371 -8.11962 -9.053900 -7.417388

Creating the cost matrix from the cost vector in the above way results in a matrix with the entries at [1,i] corresponding to the costs of being in the peak state for data point i and the entries at [2,i] corresponding to the costs of being in the background