Msgl
Original author(s) | Martin Vincent |
---|---|
Developer(s) | Martin Vincent |
Initial release | 2013 |
Stable release | 2.3.9
/ 8 May 2019 |
Written in | R, C++ |
Engine | |
Operating system | Linux, Windows, macOS, Solaris |
Type | High Dimensional Multiclass classification |
License | GNU GPLv3 |
Website | cran |
Search Msgl on Amazon.
msgl is machine learning software for high dimensional multiclass classification[1]. The optimization engine sglOptim is written in C++ and optimized for high dimensional problems with uses of multiple cores and sparse matrices for memory efficiency. The R-package msgl is a model and user friendly api wrapping of sglOptim, both developed by Martin Vincent at Roasted.Space[2] and from 2018 and onward maintained by Niels Richard Hansen[3]
Examples[edit]
Cross validation a high dimensional multiclass classification problem using 8 cores:
cl <- makeCluster(8)
registerDoParallel(cl)
v <- msgl::cv(
x = x,
classes = y,
fold = 10,
alpha = 0,
lambda = 0.01,
standardize = TRUE,
use_parallel = TRUE
)
stopCluster(cl)
Summarizing the cross validation results. The algorithm fitted 100 models with increasing number of features selected:
v
##
## Call:
## msgl::cv(x = x, classes = y, alpha = 0, standardize = TRUE, lambda = 0.01,
## fold = 10, use_parallel = TRUE)
##
## Models:
##
## Index: Lambda: Features: Parameters: Error:
## 1 1.000 1.2 20.4 0.985
## 20 0.413 21 357 0.232
## 40 0.163 29.5 501.5 0.146
## 60 0.064 30.7 521.9 0.075
## 80 0.025 30.6 520.2 0.057
## 100 0.010 31.2 530.4 0.062
##
## Best model:
##
## Index: Lambda: Features: Parameters: Error:
## 82 0.023 30.7 521.9 0.055
##
References[edit]
- ↑ Vincent, Martin; Hansen, Niels Richard (2014-03-01). "Sparse group lasso and high dimensional multinomial classification". Computational Statistics & Data Analysis. 71: 771–786. arXiv:1205.1245. doi:10.1016/j.csda.2013.06.004. ISSN 0167-9473.
- ↑ "Roasted.Space".
- ↑ "Niels Richard Hansen Homepage".
This article "Msgl" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Msgl. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.