Implements the AdaSampling procedure, a framework for both positive unlabeled learning and learning with class label noise, which wraps around a traditional classifying algorithm. See our publication for details, documentation and examples.
Yang, P., Liu, W., Yang, J. (2017). Positive unlabeled learning via wrapper-based adaptive sampling. Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, 3273-3279. [fulltext]
To install this package, use:
devtools::install_github("dinukap/AdaSampling", build_vignettes = TRUE)
library(AdaSampling)
Current version of this package includes two functions:
adaSample()
applies the AdaSampling procedure to reduce noise in the training set, and subsequently trains a classifier from the new training set.adaBenchmark()
which allows the performance of the AdaSampling procedure (with an SVM classifier) to be compared against the performance of the SVM classifier on its own.
In order to see demonstrations of these two functions, see:
browseVignettes("AdaSampling")