disr: Variable Screening With Diversity Induced Self-Representation
In abnormally-distributed/cvreg: Cross Validation and Robust Estimation Utilities

Description Usage Arguments Value References

DISR is a method of screening variables which selects variables which balance being similar to other features in a set of variables and being unique relative to other features. The balance is controlled by the user via two parameters, 'self' and 'div'.

1	disr(x, nvar = 3, self = 1, div = 1, gamma = 1.2, max_iter = 1000, tol = 1e-09)

`x`	a matrix or data frame of numeric covariates.
`nvar`	number of variables to retain
`self`	a positive number which will be used to determine how heavily self-representation is used to select variables. higher values relative to 'div' results in selection of features that are more similar.
`div`	a positive number which will be used to determine how heavily non-similarity to other features is used to select variables. higher values relative to 'self' results in selection of features that are more distinctive.
`gamma`	the parameter that controls the learning rate. defaults to 1.2.
`max_iter`	maximum number of iterations.
`tol`	convergence tolerance

a list of the loading matrix of selection indicators and a reduced 'x'.

Liu, Y., Liu, K., Zhang, C., Wang, J., & Wang, X. (2017). Unsupervised feature selection via Diversity-induced Self-representation. Neurocomputing, 219, 350–363. doi:10.1016/j.neucom.2016.09.043

abnormally-distributed/cvreg documentation built on May 3, 2020, 3:45 p.m.