A Splicing Approach to the Inverse Problem of L0 Trend Filtering

Trend filtering is a widely used nonparametric method for knot detection. This package provides an efficient solution for L0 trend filtering, avoiding the traditional methods of using Lagrange duality or Alternating Direction Method of Multipliers algorithms. It employ a splicing approach that minimizes L0-regularized sparse approximation by transforming the L0 trend filtering problem. The package excels in both efficiency and accuracy of trend estimation and changepoint detection in segmented functions. References: Wen et al. (2020) ; Zhu et al. (2020); Wen et al. (2023) .

A Splicing Approach to $\ell_0$ Trend Filtering Using Inverse Transformation

Introduction

This package provides an efficient solution for $\ell_0$ Trend Filtering, avoiding the traditional methods of using Lagrange duality or ADMM algorithms. It employ a splicing approach that minimizes L0-regularized sparse approximation by transforming the $\ell_0$ Trend Filtering problem.

R Package Installation

L0TFinv can be installed from Github as follows:

if(!require(devtools)) install.packages('devtools')
library(devtools)
install_github("C2S2-HF/InverseL0TF", repos = NULL, type = "source")

Alternatively, you can run the following code in R to install L0TFinv after downloading L0TFinv_0.1.0.tar.gz.

install.packages("Your_download_path/L0TFinv_0.1.0.tar.gz", repos = NULL, type = "source")

Usage

For a tutorial, please refer to L0TFinv's Vignette.