compboost

compboost: Modular Framework for Component-Wise Boosting - Published in JOSS (2018)

https://github.com/schalkdaniel/compboost

Keywords

Repository

C++ implementation and R API for componentwise boosting

README.Rmd

---
output: github_document
---



```{r, include=FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/"
)

library(compboost)
ggplot2::theme_set(ggthemes::theme_tufte())
ggplot = function(...) ggplot2::ggplot(...) + scale_color_brewer(palette = "Set1")

set.seed(31415)
```

# compboost: Fast and Flexible Component-Wise Boosting Framework 

[![R-CMD-check](https://github.com/schalkdaniel/compboost/workflows/R-CMD-check/badge.svg)](https://github.com/schalkdaniel/compboost/actions)
[![codecov](https://codecov.io/gh/schalkdaniel/compboost/branch/main/graph/badge.svg?token=t3Xxoxz1T2)](https://codecov.io/gh/schalkdaniel/compboost)
 [![License: LGPL v3](https://img.shields.io/badge/License-LGPL_v3-blue.svg)](https://www.gnu.org/licenses/lgpl-3.0) [![CRAN_Status_Badge](http://www.r-pkg.org/badges/version/compboost)](https://cran.r-project.org/package=compboost) [![status](http://joss.theoj.org/papers/94cfdbbfdfc8796c5bdb1a74ee59fcda/status.svg)](http://joss.theoj.org/papers/94cfdbbfdfc8796c5bdb1a74ee59fcda)

[Documentation](https://danielschalk.com/compboost/) |
[Contributors](CONTRIBUTORS.md) |
[Release Notes](NEWS.md)

## Overview

Component-wise boosting applies the boosting framework to
statistical models, e.g., general additive models using component-wise smoothing
splines. Boosting these kinds of models maintains interpretability and enables
unbiased model selection in high dimensional feature spaces.

The `R` package `compboost` is an alternative implementation of component-wise
boosting written in `C++` to obtain high runtime
performance and full memory control. The main idea is to provide a modular
class system which can be extended without editing the
source code. Therefore, it is possible to use `R` functions as well as
`C++` functions for custom base-learners, losses, logging mechanisms or
stopping criteria.

For an introduction and overview about the functionality visit the [project page](https://schalkdaniel.github.io/compboost/).

## Installation

#### Developer version:

```r
devtools::install_github("schalkdaniel/compboost")
```

## Examples

The examples are rendered using compboost `r packageVersion("compboost")`.

The fastest way to train a `Compboost` model is to use the wrapper functions `boostLinear()` or `boostSplines()`:
```{r, results="hide", warning=FALSE, fig.width=10, fig.height=2, out.width="100%", }
cboost = boostSplines(data = iris, target = "Sepal.Length",
  oob_fraction = 0.3, iterations = 500L, trace = 100L)

ggrisk = plotRisk(cboost)
ggpe = plotPEUni(cboost, "Petal.Length")
ggicont =  plotIndividualContribution(cboost, iris[70, ], offset = FALSE)

library(patchwork)

ggrisk + ggpe + ggicont
```

For more extensive examples and how to use the `R6` interface visit the [project page](https://danielschalk.com/compboost/articles/getting_started/use_case.html).

## mlr learner

Compboost also ships an [`mlr3`](https://mlr3.mlr-org.com/) learners for regression and binary classification which can be used to apply `compboost` within the whole [`mlr3verse`](https://mlr3.mlr-org.com/):

```{r}
library(mlr3)

ts = tsk("spam")
lcboost = lrn("classif.compboost", iterations = 500L, bin_root = 2)
lcboost$train(ts)
lcboost$predict_type = "prob"
lcboost$predict(ts)

# Access the `$model` field to access all the `compboost` functionality:
plotBaselearnerTraces(lcboost$model) +
  plotPEUni(lcboost$model, "charDollar")
```

## Save and load models

Because of the usage of `C++` objects as backend, it is not possible to use `R`s `save()` method to save models. Instead, use `$saveToJson("mymodel.json")` to save the model to `mymodel.json` and `Compboost$new(file = "mymodel.json")` to load the model:

```{r, eval=FALSE}
cboost = boostSplines(iris, "Sepal.Width")
cboost$saveToJson("mymodel.json")

cboost_new = Compboost$new(file = "mymodel.json")

# Save the model without data:
cboost$saveToJson("mymodel_without_data.json", rm_data = TRUE)
```
```{r, include=FALSE}
file.remove("mymodel.json", "mymodel_without_data.json")
```

## Benchmark

- A small benchmark was conducted to compare `compboost` with [`mboost`](https://cran.r-project.org/web/packages/mboost/index.html). For this purpose, the runtime behavior and memory consumption of the two packages were compared. The results of the benchmark can be read [here](https://github.com/schalkdaniel/compboost/tree/master/benchmark).
- A bigger benchmark with adaptions to increase the runtime and memory efficiency can be found [here](https://doi.org/10.1080/10618600.2022.2116446).

## Citing

To cite `compboost` in publications, please use:

> Schalk et al., (2018). compboost: Modular Framework for Component-Wise Boosting. Journal of Open Source Software, 3(30), 967, https://doi.org/10.21105/joss.00967

```
@article{schalk2018compboost,
  author = {Daniel Schalk, Janek Thomas, Bernd Bischl},
  title = {compboost: Modular Framework for Component-Wise Boosting},
  URL = {https://doi.org/10.21105/joss.00967},
  year = {2018},
  publisher = {Journal of Open Source Software},
  volume = {3},
  number = {30},
  pages = {967},
  journal = {JOSS}
}
```

## Testing

### On your local machine

In order to test the package functionality you can use devtools to test the package on your local machine:

```{r, eval=FALSE}
devtools::test()
```

Owner

JOSS Publication

compboost: Modular Framework for Component-Wise Boosting

Published

October 12, 2018

DOI

10.21105/joss.00967

Volume 3, Issue 30, Page 967

Authors

Daniel Schalk
Department of Statistics, LMU Munich

Janek Thomas
Department of Statistics, LMU Munich

Bernd Bischl
Department of Statistics, LMU Munich

Editor

Roman Valls Guimera

Tags

machine learning boosting

View PDF Review Thread Software Archive

GitHub Events

Committers

Last synced: 7 months ago

Issues and Pull Requests

Last synced: 6 months ago