Sparse Kernel Methods (SKM)

SKM is an R package for machine learning that includes functions for model training, tuning, prediction, metrics evaluation and sparse kernels computation. The main goal of this package is not to provide a full toolkit for data analysis but focus specifically in 7 supervised type of models: 1) generalized boosted machines, which internally uses gbm package, 2) generalized linear models of glmnet package, 3) support vector machines of e1071 package, 4) random forest of randomForestSRC package, 5) bayesian regression models of BGLR package, 6) deep neural networks of keras package and 7) partial least squares, which uses pls package.

The model functions in SKM were designed keeping in mind simplicity, so the parameters, hyperparameters and tuning specifications are defined directly when calling the function and the user only have to see one example to understand how the package works. The most important hyperparameters of each model can be tuned with two different methods, grid search and Bayesian optimization.

The SKM paper can be consulted in the following link: https://doi.org/10.3389/fgene.2022.887643.

Installation

Currently this package is only available in Github and as it requires the most updated versions of some dependencies, you have to make sure you have these updated versions installed. You can use the following code to install SKM:

if (!require("devtools")) {
  install.packages("devtools")
}
devtools::install_github("gdlc/BGLR-R")
devtools::install_github("rstudio/keras")
devtools::install_github("rstudio/tensorflow")
devtools::install_github("brandon-mosqueda/SKM")

If you find problems installing the library, specially windows users that does not have installed Rtools, you can use the installation script of this repository:

source("https://raw.githubusercontent.com/brandon-mosqueda/SKM/main/install.R")

Authors

• Osval Antonio Montesinos López (author)
• Brandon Alejandro Mosqueda González (author, maintainer)
• Abel Palafox González (author)
• Abelardo Montesinos López (author)
• José Crossa (author)

Citation

First option, by paper:

@article{10.3389/fgene.2022.887643,
  author={Montesinos López, Osval Antonio and
    Mosqueda González, Brandon Alejandro and
    Palafox González, Abel and
    Montesinos López, Abelardo and
    Crossa, José},
  title={A General-Purpose Machine Learning R Library for
    Sparse Kernels Methods With an Application for
    Genome-Based Prediction},
  journal={Frontiers in Genetics},
  volume={13},
  year={2022},
  url={https://www.frontiersin.org/articles/10.3389/fgene.2022.887643},
  doi={10.3389/fgene.2022.887643},
  issn={1664-8021}
}

Second option, by package:

citation("SKM")

## To cite package 'SKM' in publications use:
##
##  Montesinos López O, Mosqueda González B, Palafox González A,
##  Montesinos López A, Crossa J (2022). _SKM: Sparse Kernels Methods_.
##  R package version 1.0.
##
## A BibTeX entry for LaTeX users is
##
##  @Manual{,
##    title = {SKM: Sparse Kernels Methods},
##    author = {Osval Antonio {Montesinos López} and
##    Brandon Alejandro {Mosqueda González} and
##    Abel {Palafox González} and
##    Abelardo {Montesinos López} and
##    José Crossa},
##    year = {2022},
##    note = {R package version 1.0},
##  }