changepointsHD (version 0.3.3)

binary_segmentation: Multiple change-point method.

Description

Estimates multiple change-points using the binary-segmentation method. This does a breadth first search and uses the specified single change-point method for each sub-search.

Usage

binary_segmentation(object, method, thresh = 0, buff = 100,
  method_params = list())
# S4 method for changepointsMod
binary_segmentation(object, method, thresh = 0,
  buff = 100, method_params = list())

Arguments

object

Corresponding changepointsMod class.

method

changepointHD method for finding single change-point.

thresh

Stopping threshold for cost comparison.

buff

Distance from edge of sample to be maintained during search.

method_params

List of additional parameters for method.

Value

An updated version of the change-point model. The update will effect: 1) An estimate for the current set of change-points. 2) The mod_list, this will correspond to all the active single change-point models generated during the binary-segmentation procedure. Acitve models correspond to models that have not been superseded by more granular models. 3) The mod_range, this corresponds to the range of observations covered by each model. It can be used to determine which models are active.

Examples

# NOT RUN {
set.seed(334)

mcp_data = read.table(system.file("extdata", "mcp.txt", package="changepointsHD"))
mcp_data = as.matrix(mcp_data)

# prox gradient black-box method
cov_est = cov(mcp_data)
init = solve(cov_est)
res_map = prox_gradient_mapping(mcp_data, init, 0.1, 0.99, 0.1, 100, 1e-20)

# prox gradient black-box ll
res_ll = prox_gradient_ll(mcp_data, res_map, 0.1)

prox_gradient_params=list()
prox_gradient_params$update_w = 0.1
prox_gradient_params$update_change = 0.99
prox_gradient_params$regularizer = 0.1
prox_gradient_params$max_iter = 1
prox_gradient_params$tol = 1e-5

prox_gradient_ll_params=list()
prox_gradient_ll_params$regularizer = 0.1

simulated_annealing_params = list()
simulated_annealing_params$buff=10

changepoints_mod = changepointsMod(bbmod=prox_gradient_mapping,
                                 log_likelihood=prox_gradient_ll,
                                 bbmod_params=prox_gradient_params,
                                 ll_params=prox_gradient_ll_params,
                                 part_values=list(init, init),
                                 data=list(mcp_data))

changepoints_mod = binary_segmentation(changepoints_mod, method=simulated_annealing,
                                       thresh=0, buff=10,
                                       method_params=simulated_annealing_params)

# }